A Practical Consideration of Scanning Helium Microscopy

A PRACTICAL CONSIDERATION OF SCANNING HELIUM MICROSCOPY

Adam Joseph Fahy, BSc (Hons.)

A thesis submitted towards the degree of Doctor of Philosophy (Physics) The University of Newcastle, Australia

December 2018

ii DECLARATION

I hereby certify that the work embodied in the thesis is my own work, conducted under normal supervision.

The thesis contains published scholarly work of which I am a co-author. For each such work a written statement, endorsed by my co-authors, attesting to my contribution to the joint work has been included.

The thesis contains no material which has been accepted, or is being examined, for the award of any other degree or diploma in any university or other tertiary institution and, to the best of my knowledge and belief, contains no material previously published or written by another person, except where due reference has been made in the text. I give consent to the final version of my thesis being made available worldwide when deposited in the University’s Digital Repository, subject to the provisions of the Copyright Act 1968 and any approved embargo.

Adam Joseph Fahy

December 2018

I hereby certify that the work embodied in this thesis contains published paper/s/scholarly work of which I am a joint author. I have included as part of the thesis this written statement, endorsed in writing by my supervisor, attesting to my contribution to the joint publication/s/scholarly work.

______

Paul Dastoor

iii ACKNOWLEDGEMENTS

“Being a writer is a very peculiar sort of a job: it’s always you versus a blank sheet of paper (or a blank screen), and quite often the blank piece of paper wins.” – Neil Gaiman

Writing these acknowledgements is a rather surreal experience, as it signals an end to the unique mix of triumph and tragedy that is a PhD. I would like to take a moment to thank the long list of people who have helped me in this undertaking – without you, none of it would have been possible.

First and foremost, my thanks to my principle supervisor Professor Paul Dastoor. I have appreciated your wisdom, your patience, and most of all your enthusiasm towards the pursuit of knowledge. Ever since I turned up at your office door as a summer scholarship student (and received an 80- minute long sermon on the virtues of helium microscopy!), you have guided my development into the scientist I am today. To my co-supervisor Dr. Xiaojing Zhou – your invaluable support and wonderful sense of humour have always been sincerely appreciated. My work has benefitted greatly from your input.

When I first started, the helium microscopy at Newcastle consisted of a single PhD student working in a dark corner of a laboratory. To that lone PhD student, the now distinguished Dr. Kane O’Donnell: you managed to take a naïve undergraduate student and introduce him to the true realities of experimental science. Your talent and commitment to your work will always be an inspiration. One of my greatest joys has been to watch the evolution of the SHeM project from this humble beginning into the fully fledged research group we are now. To the members of our team, both past and present – Joel Martens, Therese Pederson, Kirren Thompson, Angus Shorter, Tom Myles, and our most recent addition, Dr. Sabrina Eder – thank

you all. Your labours have been many and varied; from the design and construction of instruments, to onerous data collection, to proof reading parts this thesis! Each and every one of you have provided much-needed comradery in face of the whims of vacuum science. I am indebted to you all, and cannot wait to see what we are able to build together in the future.

Any discussion of the SHeM team would be incomplete without mention of my dear friend and PhD brother in arms, Dr. Matthew Barr. It has been a rare privilege to undertake these long years with a trusted ally there every step of the way. My own growth a scientist has been bolstered by your remarkable acumen, mechanical expertise, and a dogged determination bordering on the obsessive. Most of all, I have appreciated your generous nature and relentless sense of humour. I cannot ever repay you for the assistance you have provided, but I look forward to trying.

The opportunity to spend three months working at the Cavendish laboratories would have been a daunting experience if not for a research group that welcomed the Australian interlopers with open arms. In particular, I would like to thank Bill Allison, Andrew Jardine, John Ellis, Donald MacLaren, David Ward, Eliza McIntosh, Barbara Lechner, David Chisnall, and Pepijn Kole. The dedication, skill and creativity on display during that period have been a constant source of inspiration during my PhD. Hopefully we can share a few more pints down at the Red Bull soon.

Finally, to my friends and family, and especially to Jasmin – words simply cannot express how much your love and support has meant to me. The patience shown through the entire endeavour has been awe-inspiring, and your ability to find the silver lining in any situation continues to amaze. Thank you for seeing me through.

v PUBLICATION LIST

The author and collaborators published the following papers during the term of the present thesis.

A. Publications related to this thesis:

 Barr, M., Fahy, A., Jardine, A., Ellis, J., Ward, D., MacLaren, D.A., Allison, W., and Dastoor, P.C. (2014). A design for a pinhole scanning helium microscope. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 340, 76-80. doi:https://doi.org/10.1016/j.nimb.2014.06.028.  Fahy, A., Barr, M., Martens, J., & Dastoor, P.C. (2015). A highly contrasting scanning helium microscope. Review of Scientific Instruments, 86(2), 023704. doi:10.1063/1.4907539.  Barr, M., Fahy, A., Martens, J., Jardine, A.P., Ward, D.J., Ellis, J., Allison, W., and Dastoor, P.C. (2016). Unlocking new contrast in a scanning helium microscope. Nature Communications, 7, 10189, doi:10.1038/ncomms10189.  Barr, M., Fahy, A., Martens, J., & Dastoor, P.C. (2016). A simple counterflow cooling system for a supersonic free-jet beam source assembly. Review of Scientific Instruments, 87(5), 053301. doi:10.1063/1.4948391.  Fahy, A., Eder, S.D., Barr, M., Martens, J., Myles, T.A., & Dastoor, P.C. (2018). Image formation in the scanning helium microscope. Ultramicroscopy, 192, 7-13. doi:10.1016/j.ultramic.2018.05.004.

B. Publications not related to this thesis:

 Fahy, A., O'Donnell, K.M., Barr, M., Zhou, X.J., Allison, W., & Dastoor, P.C. (2011). Development of an improved field ionization detector incorporating a secondary electron stage. Measurement Science and Technology, 22(11), 115902.  Barr, M., O'Donnell, K.M., Fahy, A., Allison, W., & Dastoor, P.C. (2012). A desktop supersonic free-jet beam source for a scanning helium microscope (SHeM). Measurement Science and Technology, 23(10), 105901.  O'Donnell, K.M., Fahy, A., Barr, M., Allison, W., & Dastoor, P.C. (2012). Field ionization detection of helium using a planar array of carbon nanotubes. Physical Review B, 85(11), 113404.

 Shearer, C.J., Fahy, A., Barr, M., Dastoor, P.C., & Shapter, J.G. (2012). Improved field emission stability from single-walled carbon nanotubes chemically attached to silicon. Nanoscale Research Letters, 7(1), 432, doi: 10.1186/1556-276X-7-432.  Shearer, C.J., Fahy, A., Barr, M., Moore, K.E., Dastoor, P.C., & Shapter, J.G. (2012). Field emission from single-, double-, and multi-walled carbon nanotubes chemically attached to silicon. Journal of Applied Physics, 111(4), 044326, doi: 10.1063/1.3687363.  Martens, J., Fahy, A., Barr, M., Jardine, A., Allison, W., & Dastoor, P.C. (2014). Development of a permanent magnet alternative for a solenoidal ion source. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 340, 85-89. doi:https://doi.org/10.1016/j.nimb.2014.07.033.  O'Donnell, K.M., Warschkow, O., Suleman, A., Fahy, A., Thomsen, L., & Schofield, S.R. (2015). Manipulating the orientation of an organic adsorbate on silicon: a NEXAFS study of acetophenone on Si(0 0 1), Journal of Physics: Condensed Matter, 27(5), 054002.  Andersen, T.R., Almyahi, F., Cooling, N.A. Elkington, D., Wiggins, L., Fahy, A., Feron, K., Vaughan, B., Griffith, M.J., Mozer, A.J., Sae-kung, C., Wallace, G.G., Belcher, W.J., & Dastoor, P.C. (2016). Comparison of inorganic electron transport layers in fully roll-to-roll coated/printed organic photovoltaics in normal geometry. Journal of Materials Chemistry A, 4(41), 15986-15996, doi: 10.1039/C6TA06746H.  Holmes, N.P., Marks, M., Kumar, P., Kroon, R., Barr, M.G., Nicolaidis, N., Feron, K., Pivrikas, A., Fahy, A., Mendaza, A.D.Z., Kilcoyne, A.L.D., Müller, C., Zhou, X., Andersson, M.R., Dastoor, P.C. & Belcher, W.J. (2016). Nano-pathways: Bridging the divide between water-processable nanoparticulate and bulk heterojunction organic photovoltaics. Nano Energy, 19, 495-510.  Tsarev, S., Collins, R.N., Fahy, A., & Waite, T.D. (2016). Reduced Uranium Phases Produced from Anaerobic Reaction with Nanoscale Zerovalent Iron, Environmental Science and Technology, 50(5), 2595- 2601, doi: 10.1021/acs.est.5b06160.  Tsarev, S., Collins, R.N., Ilton, E.S., Fahy, A., & Waite, T.D. (2016). The short-term reduction of uranium by nanoscale zero-valent iron (nZVI): role of oxide shell, reduction mechanism and the formation of U(v)-carbonate phases, Environmental Science: Nano, 4(4), 1304-1313, doi: 10.1039/C7EN00024C.  Almyahi, F., Andersen, T.R., Cooling, Holmes, N.P., Fahy, A., Barr, M.G., Kilcoyne, D., Belcher, W.J., & Dastoor, P.C. (2018). Optimization, characterization and upscaling of aqueous solar nanoparticle inks for organic photovoltaics using low-cost donor:acceptor blend. Organic Electronics, 52, 71-78.

vii  Holmes, N.P., Marks, M., Cave, J.M., Feron, K., Barr, M.G., Fahy, A., Sharma, A., Pan, X., Kilcoyne, A.L.D., Lewis, D.A., Andersson, M.R., van Stam, J., Walker, A.B., Moons, E., Belcher, W.J. & Dastoor, P.C. (2018). Engineering two-phase and three-phase microstructures from water-based dispersions of nanoparticles for eco-friendly polymer solar cell applications. Chemistry of Materials, 30(18), 6521-6531, doi: 10/1021/acs.chemmater.8b03222.  Pan, X., Sharma, A., Gedefaw, D., Kroon, R., de Zerio, A.D., Kilcoyne, A.L.D., Barr, M.G., Fahy, A., Marks, M., Zhou, X., Belcher, W.J., Dastoor, P.C., and Andersson, M.R. (2018). Environmentally friendly preparation of nanoparticles for organic photovoltaics. Organic Electronics, 59, 432-440, doi: 10/1016/j.orgel.2018.05.040.

viii

ix CONTENTS

DECLARATION ...... III

ACKNOWLEDGEMENTS ...... IV

PUBLICATION LIST ...... VI

CONTENTS ...... X

ABSTRACT ...... XIII

LIST OF FIGURES ...... XV

LIST OF TABLES ...... XXIX

LIST OF ABBREVIATIONS ...... XXXI

CHAPTER 1: INTRODUCTION ...... 1 1.1. Background ...... 1 1.2. Helium Atom Scattering...... 4 1.3. Contrast Mechanisms...... 7 1.4. Elements of a Neutral Atom Microscope ...... 16 1.5. Neutral Helium Imaging ...... 25 1.6. Thesis Overview ...... 27

CHAPTER 2: THE MARK I PROTOTYPE SHEM - DESIGN OVERVIEW...... 29 2.1. Development Process ...... 30 2.2. Design and Experimental Setup ...... 39 2.3. Conclusions ...... 56

CHAPTER 3: THE MARK I PROTOTYPE SHEM - EXPERIMENTAL RESULTS ...... 57 3.1. Source Characterisation ...... 57 3.2. Effusive Beam Contribution ...... 60 3.3. Optimised SHeM Imaging ...... 65 3.4. Shadowing and Masking ...... 69 3.5. Experimental vs. Model Comparisons ...... 78 3.6. Materials Studies ...... 83 3.7. Discussion and Future Work ...... 90

CHAPTER 4: THE MARK II SHEM - DESIGN AND PERFORMANCE ...... 95 4.1. Design Principles ...... 95 4.2. System Overview ...... 96 4.3. Instrument Construction...... 101 4.4. Path Length ...... 119 4.5. Critical Performance ...... 122 4.6. Conclusions ...... 125

CHAPTER 5: THE MARK II SHEM - EXPERIMENTAL RESULTS ...... 127 5.1 Image Formation ...... 128 5.2 Effusive Beam Effects ...... 139 5.3 Contrast Mechanisms ...... 148 5.4 Instrument Optics ...... 164 5.5 Conclusions ...... 184

CHAPTER 6: THE FUTURE FOR SCANNING HELIUM MICROSCOPY ...... 185 6.1 Alternate Scattering Geometries ...... 186 6.2 Contrast Specific Requirements ...... 193 6.3 Conclusions ...... 202

REFERENCES ...... 203

xii

ABSTRACT

The established field of Helium Atom Scattering (HAS) has long made use of neutral helium to offer unique opportunities with regards to surface characterisation. A thermal helium atom is an ideal probe particle: strictly surface sensitive, totally inert, a wavelength of the order of typical crystallographic dimensions, and well matched in both energy and momentum to dynamic surface processes. Technological limitations have restricted HAS to broad illumination of a sample surface. The development of a spatially resolved version of the technique - a Scanning Helium Microscope or SHeM - forms the basis for the work presented in this thesis. Such an instrument would prove of great benefit to the wide range of samples (including delicate adsorbate structures, organic molecules and biological materials) which suffer damage under the energetic probes of traditional microscopies.

Chapter 1 first reviews the nature of the helium atom-surface interaction (and the possible contrast mechanisms that arise as a result), before looking at the intensity constraints that have prevented the manufacture of a SHeM previously. Chapters 2 and 3 concern the development of a prototype instrument – the Mark I SHeM. A detailed discussion of the design decisions is included, followed by experimental studies conducted with the new instrument. With the successes found with the prototype, progress then began on creating an instrument from the ground up. Chapter 4 covers the design of the Mark II SHeM, as well as the performance improvements as compared to its predecessor. The experimental investigations into not only samples but the technique itself are explored in Chapter 5. These include studies of image formation, secondary beam effects, contrast mechanisms, and fundamental instrument optics. Finally, Chapter 6 comprises a review of the state of the emerging field with a particular focus on the technical requirements to more fully harness each of the available contrast mechanisms.

xiii

xiv

LIST OF FIGURES

CHAPTER 1 ...... 1 Figure 1.1 Examples of sample damage and image artefacts in high resolution microscopies. (a) The result of scanning electron microscope (SEM) imaging of a ternary blend film consisting of poly(3-hexylthiophene), phenyl-C61-butyric-acid-methyl-ester, and squaraine. All three organic components are known to suffer damage under the energetic probe, as indicated by the dark rectangles in the micrograph. Scale bar is 20um. (b) and (c) show the transmission electron microscope (TEM) and scanning transmission x-ray microscope (STXM) images (respectively) of a matched region of polymer nanoparticles on a silicon nitride window. Whilst the micrographs show that the nanoparticles maintain their structure under the electron and x-ray beams, spectroscopic STXM studies revealed their chemical structure had been permanently altered through the breaking of carbon-carbon double bonds. (d) and (e) show SEM micrographs of silver nanowires and a polyacetonitrile film respectively. Both images display the edge enhancement and charging effects that can occur during SEM imaging. Scale bars are 1um and 10um respectively. Micrographs courtesy of Natalie Holmes...... 2 Figure 1.2 Comparison of the wavelength as a function of probe energy for photons, electrons, neutrons and helium atoms...... 3 Figure 1.3 Possible scattering mechanisms for a thermal helium atom incident on a surface. Dotted lines represent equipotentials for the helium-surface interaction. Trajectory 1 – elastic scattering, yielding a diffraction pattern characteristic of the surface periodicity. Trajectory 2 – inelastic scattering as a result of energy exchange with phonons or adsorbates. Trajectory 3 – temporary trapping of the impinging atom into a resonant state of the interaction potential. Permanent trapping (adsorption) of the atom is also possible, but very unlikely for thermal helium. Figure courtesy of Barr [32], originally adapted from Toennies [33] and MacLaren [24]...... 6 Figure 1.4 Schematic illustrating the classification of the contrast mechanisms for scanning helium microscopy. The contrast available to the technique ultimately stems from the nature of the probe-sample interaction – at the highest level, an elastic or inelastic interaction. From these scattering trajectories, we may define our three contrast mechanisms: topological, diffractive, and chemical (yellow text)...... 8 Figure 1.5 Geometry for the 2D model for topological contrast from a surface with perfect diffuse elastic scattering. The diagram shows two planes (red) inclined by an angle ±δ from a mean plane whose normal (arrow) is itself inclined at an angle θ with respect to the detector direction (green)...... 11 Figure 1.6 Schematic drawing of isolated (low-coverage) CO molecules on a Pt(111) surface. The solid line represents the corrugation function (‘repulsive wall’) for a room temperature helium beam. The dashed and dotted semi-circles correspond to the cross-sections of adsorbed CO (ΣCO) and gas phase CO (σCO) respectively. The large cross-sections for helium scattering from these adsorbates are also given, illustrating the sensitivity of a helium beam to individual adatoms. Figure from [25]...... 15 Figure 1.7 The shape of the effusive beam caused by molecular flow through a channel of length L and radius r can be determined by considering the Knudsen number K and the parameter β...... 17 Figure 1.8 Schematic representation of a supersonic free-jet expansion. A dense volume of gas is allowed to expand through a small nozzle into a region of lower pressure. The high numbers of interatomic collisions in the initial turbulent flow gives rise to a supersonic expansion outwards from the nozzle. The rapidly decreasing density leads to the flow regime transitioning from fluid flow to free-molecular, with the boundary between traditionally termed the ‘quitting surface’. Past this point in the expansion (the ’zone of silence’), the gas particles are considered to have a large mean free path and travel along straight-line trajectories. It is typical to sample the centre of the expansion through the insertion of a sharp conical aperture or skimmer into the zone of silence. Schematic adapted from [24]...... 18

xv Figure 1.9 A summary of the potential methods for focusing neutral atoms to a point (Image courtesy of [32], adapted from MacLaren [24])...... 19 Figure 1.10 SEM micrograph of a section of a free-standing silicon nitride Fresnel zone plate used to focus helium beams. The inset shows the outermost zone with an approximate periodicity of 96 nm, demonstrating the technical skill necessary to create structures capable of interacting as desired with neutral helium atoms. Image courtesy of Reisinger et. al. [58]...... 21 Figure 1.11 The Cambridge Spin-Echo Neutral Helium Detector, demonstrating the necessary increase in size to compensate for the low efficiency of the electron ionisation process. Image courtesy of David Chisnall...... 23 Figure 1.12 SEM image of a typical electrochemically etched tungsten tip with end radius of the order of 30 nm...... 24 Figure 1.13 The first neutral helium images as produced by Koch et. al. [58] of a hexagonal TEM grid. Micrograph (a) was produced with a beam focused down to a 3 micron spot and 8 seconds collection per pixel, while the zoomed region (b) used a 2 micron spot and 14 seconds collection time per pixel...... 26 Figure 1.14 Schematic view of the instrument geometry for the neutral atom microscope. In the most recent iterations, the nozzle is held between 300 and 600 microns from the pinhole aperture, while the working distance is typically between 10 and 50 microns. Together, the minimisation of the distance from source to sample enables the instrument to generate a large helium flux incident on the sample surface. Image courtesy of [87]...... 26

CHAPTER 2 ...... 29 Figure 2.1 Sample simulated micrographs for two materials with 10% contrast and a count rate of 1000 Hz. From left to right, the number of pixels in the sample images are as follows: 16x16, 32x32, 64x64, 128x128, and 256x256. As the pixel count increases the proportional effect of shot noise is diminished, resulting in a clearer distinction between the materials...... 30 Figure 2.2 Sample 64 x 64 pixel images produced for a range of count rates and material contrasts. In the planning process for the prototype, it was decided that the build would go ahead if an image of a quality greater than a set level could be produced, represented by the images to the right of the dotted line in the above figure...... 31 Figure 2.3 A traditional helium atom scattering system; in particular the apparatus at Bell Labs as described by R. B. Doak in [41]. Designs such as this were the starting point for the layout of the prototype SHeM to be built at the Cavendish laboratories...... 32 Figure 2.4 Schematic diagram of the prototype SHeM. The helium beam source consists of a free-jet expansion, of which the centreline is selected out with a skimmer in the source chamber. The beam passes through a differential pumping stage to the pinhole optics of the instrument, a 5 micron FIB milled pinhole. The result is a thin beam of helium striking the sample surface in the sample chamber, with the scattered helium entering the detector chamber where it stagnates to form a stable pressure. By rastering the sample back and forth under the beam, an image of the surface may be constructed...... 35 Figure 2.5 Representation of how the gas flow model handles the interaction of the helium beam with the sample surface. The same number of helium atoms incident on the sample from the free-jet beam are then spread evenly into a hemispherical cap centred where the beam strikes the sample...... 37 Figure 2.6 The previous vacuum system at the Cavendish laboratories, part of which was used to build the prototype SHeM. In particular, the source chamber and section of frame it rests on was repurposed for the new instrument...... 39

xvi

Figure 2.7 Front cross-section of the source for the prototype SHeM, with inset of more detail of the nozzle assembly. The source chamber (green) and Shimadzu TMU2203 turbomolecular pump (brown) were reused from a previous supersonic beam source, as was the welded frame they sat upon. To form the helium source for the prototype SHeM, a new nozzle assembly (red) was constructed in the style of Buckland et. al. [55] with a 10 um nozzle fed by stainless steel pipework (not shown). The nozzle assembly was mounted to an X-Y-Z UHV Designs manipulator (blue) in order to precisely control the position of the nozzle with relation to the Beam Dynamics skimmer (orange)...... 41 Figure 2.8 Photos of the nozzle assembly during construction. Note the designated axes for movement of the nozzle...... 42 Figure 2.9 Sectioned exploded schematic of the DN160CF flange on which the skimmer is mounted. The base of the clamp can be swapped out to change the travel range for the source nozzle relative to the skimmer...... 42 Figure 2.10 Photograph of the mounted skimmer prior to installation into the SHeM...... 43 Figure 2.11 Top cross-section of the prototype SHeM source (green), differential and sample chambers (blue). The differential stage consists of a DN40 tee welded to the front of the sample chamber flange, allowing for a turbo pump to be connected via bellows. The beam passes through the differential stage and onto the sample chamber by passing through the pinhole plate (purple), wherein the pinhole apertures out all but the desired spot...... 44 Figure 2.12 Illustration of the potential ways to implement a pinhole at 45 degrees to the beam axis (represented by the green arrow). If the pinhole is added normal to the plane of the membrane as in (a), then unless the membrane is thinner than the width of the pinhole line of sight to the sample is lost. In (b) the pinhole is bored along the beam axis, but manufacturing and alignment concerns make this difficult. (c) displays the compromise utilised in the prototype SHeM: A pinhole is bored through a very thin section of material, allowing the beam to pass through as desired. In order to implement this practically, the pinhole substrate was chosen to be a silicon nitride disc with a small membrane section in the center, thus allowing it to handled and glued into position...... 45 Figure 2.13 Sectioned render of the pinhole plate used to hold the silicon nitride disc with the pinhole bored through the center. The channel to the right shows the beam entry pathway, including the shallow depression on the face of the plate where the silicon nitride disc is located. The second pathway is to allow reflected helium to travel on to the detector chamber. The point where the axes of both channels meet (as shown in the inset) is the sample specular position...... 46 Figure 2.14 Photograph and optical micrographs of the 5 um pinhole implemented in the prototype SHeM. The silicon nitride disc was glued in place in the pinhole plate with vacuum leak sealant to provide a strong bond and stop extraneous gas escaping around the edges of the disc. As can be seen in the micrographs, the 250 micron square film has the pinhole bored through its center. The rippling of the membrane was caused by the field ion beam milling process...... 47 Figure 2.15 Clockwise from top left: Sample slide with Z-distance screw in place and a TEM grid mounted, ready for imaging; Sample mount mechanism assembled out of the chamber (note the designated labels for the movement of the sample); one of the Attocube ECS3030 piezo slipstick drives responsible for the rastering of the sample underneath the beam; the sample mount mechanism assembled and bolted to the pinhole plate, showing the typical imaging position...... 48 Figure 2.16 Sectional top view of the sample mount (red) along with the pinhole plate (purple) as mounted to the inner wall of the sample chamber (blue). The incident beam axis is marked in dotted lines, with the specular position (and hence the sample location) at the point where it crosses the axis to the detector...... 49 Figure 2.17 Cross-sectional side view of the sample and detector chambers highlighting the path of the reflected helium atoms from the sample through to the Hiden quadrupole. A 1 mm diameter aperture in the pinhole plate admits a portion of the specularly reflected helium atoms, which then pass through the custom sample chamber flange, onto the detector chamber proper. A stable population builds, which is sampled by the quadrupole...... 50

xvii Figure 2.18 Photograph from the side of the prototype instrument in operation, with the source chamber on the left and the sample chamber with its turbo pump to the right. Also visible is the cart the source chamber sits on, allowing it to be moved back from the rest of the chambers to provide access for skimmer interchanges and adjustments to the nozzle assembly...... 52 Figure 2.19 Photos of the completed instrument during operation. The top photo shows a top down view of the beam path from source to sample to detector (the latter wrapped in foil for baking purposes at the time). The bottom photo looks back past the detector chamber towards the sample and source chambers...... 53 Figure 2.20 Photograph of the front of the beam control panel, showing the regulators, valves and gauges used when supplying the compressed helium to the nozzle. Behind the panel face sits the Swagelok pipework (see schematic below). Additionally. you can see the helium cylinder to the left, the gas booster at the bottom center, and the dewar which comprises the outer shell of the cold trap to the bottom right...... 54 Figure 2.21 Gas flow schematics of the beam control panel and prototype SHeM. High vacuum stages are shown in black, rough vacuum in blue, high pressure components in red, and gas storage in green. The gas panel allows bottled helium to be compressed to the neighbourhood of 200 bar and then regulated down to the desired stagnation pressure. It should also be noted that the helium was passed through a cold trap in order to filtrate any remaining impurities. Bypasses link the chambers and thus prevent the possibility of damage to the skimmer or pinhole when bringing the system up to atmosphere (or the reverse) ...... 55

CHAPTER 3 ...... 57 Figure 3.1 Plot of the corrected source chamber pressure as a function of beam stagnation pressure for the prototype SHeM. The linearity of this data, as shown by the quality of the linear fit (R2 value of 0.997), demonstrates the expected performance for a supersonic free-jet beam source. From the slope of the fit – (3.81 ± 0.18) x 10-5 – we derive an estimate for the effective nozzle diameter of (10.08 ± 0.24) microns, in good agreement with the nominal diameter of 10 microns...... 58 Figure 3.2 Helium signal as detected by the Hiden quadrupole as a function of nozzle-to- skimmer separation both with and without the sample stud sitting at the specular position. Scans were performed with a stagnation pressure of 120 bar at room temperature, and a pinhole plate without pinhole installed between the differential and sample chambers. The difference in signal is due to the helium beam reflecting off the metal surface into the detector aperture...... 59 Figure 3.3 Theoretical angular distributions for effusive beams under different conditions. (a) Angular dependence of the beam profile on the value of β for a source with a Knudsen number K >> 1. β = ∞ refers to the case of a thin walled orifice, which then causes the effusive beam to follow the familiar cosine distribution. (b) Demonstrates the change when K ≤ 10, and the shape becomes dependant on both K and β. For the given β value of 0.05, the directivity of the beam source can be seen to increase with larger Knudsen numbers. Figure from [51] ...... 61 Figure 3.4 Detected helium signal as a function of varying nozzle-to-skimmer separations. At small separations, the pressure in the differential chamber will increase such that the effusive beam will dominate the free-jet beam, producing a broad peak unsuitable for imaging. However, by pulling the nozzle back from the skimmer the differential stage pressure drops and the sharp free-jet beam emerges. Note that the nozzle-to-skimmer separations have a zero offset of up to 1 mm due to the difficulty of aligning the nozzle with the fragile skimmer...... 63 Figure 3.5 Comparison of the differential stage pressures as nozzle position is varied at the extremes of the nozzle-to-skimmer separation from Figure 3.4. At 10mm separation, the helium partial pressure is an order of magnitude smaller than that at the 5 mm separation (1.4 x 10-5 mBar vs. 1.7 x 10-4 mBar in the central position)...... 63 Figure 3.6 The SHeM micrograph on the left shows a region of a TEM grid before the nozzle- to-skimmer distance was increased, while that to the right was taken afterwards. The appearance of the expected grid pattern with the increase in distance demonstrates the clear difference in instrument performance once the secondary effusive beam contribution is minimised...... 65

xviii

Figure 3.7 Scanning helium micrograph (left) and matched reflection optical micrograph (right) of a region of TEM grid mounted using carbon tape on a sample stud. TEM grid has a bar width of 24 um, and a pitch of 84 um. Under the helium beam, the carbon tape appears as bright as the copper grid, while the stainless steel of the stud is much darker...... 65 Figure 3.8 SHeM micrographs of a TEM grid demonstrating the problems with drift in the horizontal axis due to a problem with the translation stage. In some instances the stage would drift initially but then work as expected (Figure 3.7 is such an example), while in others it might continually drift throughout the entire scan (left image). The most detrimental type of drift was an inconsistent variation throughout the scan (right image), making post-correction difficult. Note in the right hand image, the transition between the bright and dark areas underneath the TEM grid correspond to the carbon tape / sample stud interface as described for Figure 3.7...... 67 Figure 3.9 Schematic depicting horizontal drift causing image artefacts. Instead of the expected scan path shown in (a), the problems with the Attocube stage controlling horizontal motion would cause it to drift further in one direction, resulting in the scan path to instead be that shown in (b)...... 68 Figure 3.10 Two SHeM micrographs of adjoining regions of a TEM grid taken in succession. The lack of closed loop control in the scan stages leads to the mismatch in horizontal starting position...... 68 Figure 3.11 Simplified 2D illustration of masking and shadowing in the SHeM. Image (a) shows a simple sample with a large asperity being scanned under a helium beam (green) as shown by the arrow. When the beam strikes the sample sufficiently far away from the asperity, it makes no impact on the collected helium signal. However, in (b) the sample has been scanned across, leading to the reflected signal from the surface striking the protrusion, meaning that an area is masked due to detector occlusion. Further on at (c), the asperity is now blocking the beam from illuminating a section of the sample surface (shadowing). Note that part of this shadowed region will include the top of the asperity – provided the detector acceptance angle is sufficient, helium striking the top will enter the detector, forming the image at this position. Finally, the asperity moves far enough to no longer impact the incident beam as shown in (d)...... 70 Figure 3.12 Trimetric projection of a simple geometric sample being imaged via SHeM. Red arrow indicates the beam position for a masked area, which results in the cross hatched region of lower intensity. The blue arrow indicates a position where the substrate is shadowed by the sample - in fact, each pixel showing the sample will constitute a shadowed region where the beam was prevented from striking the substrate. Image to the right displays the resultant micrograph...... 71 Figure 3.13 Illustration of the consequences of the incident beam geometry on the produced SHeM micrographs. The helium beam enters from the right at an angle α relative to the normal for the base plane (for the Mk I SHeM, α=45º), and interacts with a simple sample (a needle). Specular reflections from the sample and base plane are shown via the red arrows, with the detector positioned to the left. Yellow indicates a shadowed region, while blue represents masking. As θ (the angle between the needle and the base plane) increases from 0º through to 180º as shown in the progression from (a) through to (e), the apparent length of the feature (shadowed region) can be seen to change significantly, a phenomena known as projection distortion. When the needle sits normal to the base plane as in (c), the 45º incident beam results in the apparent length equalling the true length. Note that once θ moves past 135º (as in (e)), the shadowed region appears to reverse direction, with the far side of the needle now exposed to the helium beam...... 72 Figure 3.14 Plot of the scaling factor SF = sin(θ) + cos(θ) in a SHeM micrograph for a feature (needle) at an angle of θ (see inset) with respect to the sample slide. A plane tilted away from the incident beam (0º < θ < 90º) will appear longer in a SHeM micrograph, while those tilted towards the beam (90º < θ < 180º) will appear shorter. Negative values of the scaling factor (135º < θ < 180º) indicate the apparent length of the plane has reversed direction (for example, would appear to the right of the base of the needle as shown in Figure 3.13 e). Note that this scaling factor will only apply in the horizontal scan axis – in the vertical direction the image will be in direct correspondence with the sample...... 74

xix Figure 3.15 Optical image of the section of TEM grid to be imaged, mounted on the sample stud (white area) with carbon tape (black). A drill bit had been used to excise the central portion of the grid to allow it to function as a locator for other samples, resulting in damaged spars being bent at a variety of angles with respect to the plane of the grid. The box indicates the region imaged in Figure 3.16...... 75 Figure 3.16 Scanning helium micrograph (left) and matched reflection optical microscope image (right) of a region of TEM grid mounted via carbon tape. The SHeM micrograph shows both masking and shadowing due to the carbon tape below the grid and the bent spars of the grid sitting above the bulk of the sample (respectively). Letters (a) through (f) correspond to features detailed in the text overleaf...... 75 Figure 3.17 Schematic cross-section of the helium trajectories interacting with the bent TEM grid. Scan positions represented by (a) and (b) will result in dark areas in the final image (as in the top right of the micrograph) due to masking and shadowing respectively, with the bulk of the helium beam unable to make it to the detector. (c) and (d) will yield bright regions in the produced micrograph, with (c) the more intense as the mean plane is tilted towards the detector, as well as being closer to the entrance aperture...... 76 Figure 3.18 Optical microscopy of a typical copper TEM grid such as that imaged first in the prototype SHeM. Detail callout shows the very rough surface of the grid, leading to diffuse scattering of the helium beam. Scale bar is 40 microns in length...... 77 Figure 3.19 Location of the three linescans across the edge of a TEM grid spar used to obtain an estimate for the resolution of the prototype microscope...... 79 Figure 3.20 Result of the averaging of three vertical line profiles across the edge of a TEM grid as imaged by the prototype SHeM. Note that the intensities have been normalised for ease of analysis. Looking at the sample travel required to move from 20% to 80% of the intensity, an instrument resolution of (3 ± 2) microns is obtained...... 80 Figure 3.21 Optical micrograph (top) and matched scanning helium micrograph (bottom) of a section of wing belonging to the fly species musca domestica. The SHeM micrograph reveals much more of the detail in the transparent surface, including the convoluted folds of the wing membrane. Note that due to the surface sensitive nature of the neutral helium probe particle, the SHeM micrograph contains only information from the top side of the wing (unlike the optical). . 83 Figure 3.22 SHeM micrograph of a section of fly wing overlaid on a composite of optical images of the same sample. Such a large image (100 x 400 pixels) required the instrument to scan for approximately 48 hours. Scale bar is 300 microns in length...... 84 Figure 3.23 Cropped section of the SHeM micrograph shown in Figure 3.21. Rescaling the colour based on the intensities in this smaller portion allows more of the detail to be visible, including the hairs attached to the top edge of the wing...... 85 Figure 3.24 Top down SEM (left) and optical micrographs (right) of the ‘Universal Resolution Sample’ available from Agar Scientific (AGS1937). SEM image from the technical data provided by Agar Scientific [109]. The sample consists of tin spheres laid down on a mechanically polished carbon substrate, with spheres sizes ranging from approximately 30 microns in size down to less than 5 nanometres. In addition to resolution and masking effects, the sample is of interest to SHeM due to the difference in the two materials (visible as the contrast in the SEM micrograph)...... 87 Figure 3.25 SHeM micrographs of the tin sphere sample. In the left image, each pixel represents a step of 1.1 microns, while on the right 1.3 microns. Note that the beam enters from the bottom (90 degree rotation as compared to previous SHeM micrographs) to guide the eye...... 87 Figure 3.26 Section of SHeM micrograph showing a single tin sphere...... 88 Figure 3.27 Micrographs of a section of polymer bonded explosive as imaged via SHeM (left) and a polarised reflection optical microscope (right). The sample in question had been mechanically polished flat to within 50 nanometres, leading to the likely source of contrast in the image being topological contrast below the resolution limit of the instrument (i.e.: roughness). Pixels in the SHeM micrograph represent a sample movement of approximately 4 microns. .... 89

Figure 3.28 Cross-section of the prototype SHeM highlighting the shape and size of the differential pumping stage. The length of the chamber (indicated by the red arrow) along with the remote placement of the turbo pump would ideally be changed in the next generation of the instrument, meaning a fundamental redesign...... 92

CHAPTER 4 ...... 95 Figure 4.1 Schematic diagram of the SHeM II system. Using a 10 μm nozzle, a supersonic free-jet expansion of neutral helium is produced in the source chamber. A skimmer samples the centerline of the expansion, which then passes into the differential stage. The optics of the instrument consists of a silicon nitride membrane with a FIB milled pinhole mounted into a metal plate (see inset). The beam is incident on the back of this membrane, leaving a small spot of helium free to strike the sample. The sample itself is able to be rastered underneath the beam via three linear drives to facilitate imaging. A portion of the helium reflected from each point on the sample surface passes through a second aperture in the pinhole plate into the detector chamber, where the stagnation pressure is sampled to yield the intensity...... 97 Figure 4.2 Photograph of the Mark II SHeM constructed at the University of Newcastle. Visible on top of the system frame is the source chamber to the left, attached to the sample chamber (door with the window in the center). Along the back wall can be seen the gas panel responsible for the compression and regulating the helium supply to 200 bar, fed into the top of the source chamber through the visible thin pipework. The black insulated gas lines are those bringing chilled nitrogen into the same source assembly...... 98 Figure 4.3 Photograph of the Mark II SHeM displaying (from left to right) the Hiden quadrupole control box, the detector chamber and pumping connections, the box chamber housing both the sample and differential stages, and finally the source chamber including helium and nitrogen gas supply lines. Note that the pipework for the chilled nitrogen gas (braided cables) have not yet been insulated in this photo (as compared to Figure 4.2 previous)...... 99 Figure 4.4 Gas flow schematic for the Mark II scanning helium microscope. High vacuum stages are shown in black, rough vacuum in blue, high pressure components in red, and gas storage in green...... 100 Figure 4.5 Cross-sectional schematic of the Mark II SHeM source chamber as viewed from the front. The source chamber (green) was pumped by a large turbo pump (Edwards STP- iXR2206) through the bottom ISO250 flange. A nozzle assembly (red) based on the style of Buckland et. al. [55] with a novel cooling system was mounted to an UHV Designs manipulator (blue) allowing precise control of the nozzle relative to the skimmer...... 102 Figure 4.6 Plot of the corrected source chamber pressure as a function of beam stagnation pressure for the Mark II microscope. Linear fit to data has an R2 value of 0.991 (indicating the source was functioning correctly) and a slope of (3.39 ± 0.25) x 10-5. From the latter, we derive an estimate for the effective nozzle diameter of 10.10 ± 0.37 microns, in good agreement with the nominal diameter of 10 microns...... 103 Figure 4.7 Schematic cross-sectional diagram of the helium beam source for the Mark II SHeM design. A Buckland-style nozzle assembly [55] with a (nominally) 10 micron aperture was constructed, capable of withstanding stagnation pressures of at least 250 bar at cryogenic temperatures. The stainless steel pipework that supplied the compressed helium to the nozzle was passed through a larger outer pipe as shown, allowing a counter current of chilled nitrogen gas to provide cooling. Copper blocks on the nitrogen inlet and surrounding the nozzle assembly, linked by a copper bridge piece, acted to stabilise the temperature. Surrounding the nozzle copper block was a heater clamp, allowing the stagnation temperature to be precisely set via a PID controller...... 104 Figure 4.8 Photograph of the source assembly for the Mark II. Surrounding the hexagonal VCR cap at the centre (holding the 10 micron aperture) is a copper block followed by a heater clamp to enable control over the temperature of the stagnation volume when used in concert with the counterflow of chilled nitrogen gas. Also visible to the top left is the secondary copper block on the nitrogen inlet pipe, as well as copper braid connecting the two cooling blocks (an initial linkage that was eventually replaced by the solid copper bridge shown in Figure 4.7)...... 105

xxi Figure 4.9 Plot of the temperatures at the top of the source chamber manipulator as well as the two copper blocks as a function of time during operation of the cooling system. Also included is the source chamber pressure, observed to increase due to the change in centreline intensity predicted by equation 3.1 with gas temperature...... 106 Figure 4.10 3D render of the skimmer mounting plate (blue) and clamping ring (red) used to secure a 100 micron ‘Model 2’ skimmer as produced by Beam Dynamics Incorporated. The mounting plate may be swapped out for other versions with different depths, thus allowing the distance from skimmer to pinhole to be varied...... 107 Figure 4.11 Top-down cross-sectional schematic of the redesigned differential pumping stage. On the left, the 8” CF flange forming the edge of the source chamber (including nozzle assembly in red) is visible. The flange joins to the side of the box chamber (brown) containing both the differential and sample chambers, separated by an internal wall. Just inside the connection, the skimmer mounting plate (see Figure 4.10) sits, aligning the skimmer to the beam axis and setting the skimmer to sample distance. The center of the expansion produced by the nozzle is selected out by the skimmer, passes through the differential stage, and enters the back of the pinhole plate via a hole in the internal wall to be incident on the silicon nitride membrane with pinhole. Reflected helium from the sample is able to pass through to the detector volume via a 1 1/3” half nipple internally welded to ensure isolation from the differential volume. Note the visible 4.5” port on the bottom of the differential stage, allowing the direct connection of an Edwards EXT75 DX turbomolecular pump to ensure maximum pumping to the volume...... 108 Figure 4.12 3D render of the connection between the source and differential chambers, with various levels of sectioning to help illustrate the complex geometry used to ensure a condensed beam path while still providing significant pumping...... 109 Figure 4.13 Top-down cross-sectional schematic of the box chamber comprising the differential and sample volumes. Note that for simplicity, the sample mount and the hinged door which forms the front wall have been omitted...... 111 Figure 4.14 3D render of the full sample mount which consists of a base plate, stand, three Attocube ECS3030 units to drive the sample in the X, Y, and Z axes as shown, and the three-pin kinematic mount into which the sample slides are placed. Inset shows a photograph of the stacked ECS3030 stages and kinematic mount with sample slide...... 112 Figure 4.15 Photograph of the inside of the sample chamber with the top level of the sample mount removed, illustrating the placement in front of the pinhole plate...... 113 Figure 4.16 Sectioned 3D drawing of the detector chamber. The Hiden quadrupole (red) sits tightly within the main body of the chamber, minimising the stagnation volume. Further in this regard, the DN16CF gate valve (green) used to isolate the detector has been incorporated directly into the sheath. A small turbomolecular pump (Edwards EXT75 DX) is connected to the main body of the chamber via two DN40CF butterfly valves (blue). By changing the extent to which each valve is opened, the pumping on the stagnation volume can be controlled by the user. . 116 Figure 4.17 3D render of the system frame, constructed from steel I-beams (red) and hollow square section (green). The entire frame was supported by three novel vibration isolators (callout). Also visible are the linear bearing rails (yellow) used to support the source chamber cart and allows access to the nozzle assembly or skimmer as required...... 117 Figure 4.18 Screenshot of the LabView user interface for the Mark II SHeM. In addition to the basic control over the current position of the sample and the parameters of the scan to be run, the program also shows the image currently being generated along with a profile for the current horizontal line. Furthermore, once a scan is complete, the datafile can be reopened in this interface and a region boxed to start a new scan. As such, imaging features of interest within a prior scan becomes very quick and easy, especially as compared to the Mark I instrument. .. 118 Figure 4.19 Illustration of the ‘locator scan’ used when first loading a new sample slide into the Mark II SHeM. An optical image (typically either an optical micrograph or, as in the left image above, a photograph) would be matched to a quick, large area SHeM scan (right image) in order to identify all major features. By selecting smaller regions of this large scan, high resolution images were produced. Scale bar is 1 mm in size...... 119 Figure 4.20 Helium beam path through the microscope geometry, broken down into smaller sections for the purpose of comparing the improvement of the Mark II over its predecessor. .. 120

xxii

Figure 4.21 Illustration of the potential extension of the pinhole plate from its current geometry (dotted line) to a point much closer to the sample surface, the effect of which is a reduced working distance and a much larger count rate...... 121 Figure 4.22 Matched optical (left) and neutral helium (right) micrographs of a copper TEM mounted to a sample slide with a central hole. SHeM micrograph produced using a 1.3 second dwell per pixel and a step size of 12.5 um. Scale bar is 200 microns in length...... 122 Figure 4.23 Detected helium signal as the nozzle is scanned across the skimmer for a series of nozzle-to-skimmer separations for the Mark II SHeM. For said experiment, a 200 bar beam at room temperature was utilised. In stark contrast to the similar plot collected for the Mark I, even with the nozzle at closest approach the free-jet beam is observed to dominate, although the presence of the effusive beam is still visible in the broad shoulders for this scan. Pulling the nozzle further back diminishes the pressure in the differential stage and hence lowers the effusive beam intensity (at the cost of some loss to the available centreline intensity). Note that the nozzle-to- skimmer separations have a zero offset of up to 1 mm due to the difficulty in aligning the nozzle to the fragile skimmer...... 123 Figure 4.24 Plot of maximum differential chamber helium partial pressure as a function of nozzle position. Helium through the skimmer orifice is inversely proportional to the square of the nozzle-to-skimmer separation, as shown by the quality of the inverse-square fit (blue dashed line, R2 = 0.999)...... 125

CHAPTER 5 ...... 127 Figure 5.1 SHeM micrograph of a hexagonal TEM grid produced with the Mark II scanning helium microscope using a 3 micron step size (scale bar is 100 microns in length). The helium beam strikes the sample from the right side of the image, with the detector aperture sitting to the left. Suspension of the grid off the substrate with carbon tape yields the strong helium shadows observed beneath the grid...... 127 Figure 5.2 Matched optical (left) and neutral helium (right) micrographs of a copper TEM grid (Ted Pella part number: 8GC90) mounted to a sample slide with a central hole. SHeM micrograph produced using a a step size of 12.5 um; scale bar is 200 microns in length. The pitch of the grid (indicated by doubled headed arrow) as measured from the micrograph is (280 ± 25) microns – well matched to the expected value of 282 microns as provided by the manufacturer...... 128 Figure 5.3 SHeM micrograph (10 micron step size, scale bar 200 microns in length) of the central portion of a Norcada silicon nitride x-ray window (part number NX5025Z), with dimensions of the window shown to the right. The projection distortion of the image along the plane of the beam and detector (horizontal plane in the SHeM micrograph) is clear. While the square outer edge of the frame remains unchanged (as a feature parallel to the base plane), the size of the inclined planes down to the actual membrane in the center are different depending on orientation. The inclined planes are steeper than the incident beam, and as such a portion of the square membrane is shadowed and thus blocked from view...... 129 Figure 5.4 Left: SHeM micrograph of a sugar crystal adhered to a carbon dot (500 micron scale bar, 6 micron step size). Right: CAD render of a replica crystal with light source and camera positioned in the same arrangement as in the SHeM. Note however that the camera and light source had to be swapped in order to produce the shadow as in the SHeM image (shadowing versus masking)...... 130 Figure 5.5 SHeM micrograph of a sugar crystal adhered to a carbon dot (500 um scale bar, 6 micron step size). The relative intensities of the various faces make sense if we consider the helium to scatter predominantly diffusely from the crystal surface. The brightness of side ‘4’ as compared to ‘3’ - despite also being obscured from the detector aperture - is due to multiple scattering events with the carbon substrate...... 131 Figure 5.6 Assuming predominantly diffuse scattering from the sample surface, the schematic illustration shows how each crystal face (labelled as in Figure 5.5) results in different amounts of helium able to make it to the detector. While face ‘3’ reflects helium back towards the incoming beam, ‘4’ has a much greater opportunity to cause multiple scattering events with the assistance of the substrate, leading to a greater intensity in the final micrograph...... 132

xxiii Figure 5.7 (a, b) SHeM micrographs of a section of TEM grid (such as shown in the optical micrograph (c)) acquired at working distances of 1.98 and 3.40 millimetres respectively. Note that the range of intensity values has been set identically for both micrographs to allow direct comparison. The masked region caused by the spar in the SHeM images (see arrows) can be observed to shift position, and details such as the shallow depression along the center of each of the TEM grid spars (diamond arrowhead in (a), and visible as a difference in shading in the optical micrograph (c)) become more evident at smaller working distances...... 133 Figure 5.8 Schematic illustrating the effect of different working distances on the scattering geometry. Moving the sample closer to the pinhole plate is achieved by varying the Z-position - by then adjusting the X-stage similarly, we can bring the same feature back into line with the beam. As the different coloured cones suggest, the alteration of working distance changes the relative position of the detector aperture, as well as its acceptance angle...... 134 Figure 5.9 SHeM micrographs of a TEM grid suspended over the edge of a section of carbon tape conducted at different working distances (calculated from known sample to pinhole separations), namely (a) 0.74, (b) 1.44, (c) 2.15, (d) 2.86 (specular), (e) 3.56, (f) 4.27, (g) 4.98, (h) 5.69, and (i) 6.39 millimetres. As the effective detector angle changes, the masked area of the sample surface by the grid spar can be seen to shift, giving a rudimentary form of 3D imaging...... 135 Figure 5.10 Plot of the magnitude of the topological contrast as given by equation 5.1 for a range of values of θ and δ. Note that when the condition that θ + δ < 90o is broken (ie: the detector line-of-sight is blocked - top right half of the plot), the contrast has been set to zero for clarity. The region between the blue dot-dashed lines indicate the range of θ values for the Mark II SHeM with the sample at specular...... 137 Figure 5.11 (a) Anaglyph of a sugar crystal as built from two SHeM micrographs as imaged using (b) a new sample mount (CAD render) for the Mark II SHeM designed by Myles [123]. The sample mount allows for two modes of 3D imaging and is currently undergoing testing at the time of writing...... 138 Figure 5.12 (a) SHeM micrograph of a TEM grid adhered to a cleaned silicon wafer with a small piece of carbon tape (20 micron step size used; scale bar represents 1 millimetre). (b) and (c) show sections of the grid – indicated by the square shown in (a) – imaged using a beam with differing ratios of supersonic to effusive beam. (b) shows the grid as imaged with the raw supersonic beam and effusive contribution native to the SHeM with a nozzle-to-skimmer separation of 11 mm, while (c) includes a more significant effusive beam contribution. To cause the latter, the differential chamber had an additional 4.0 x 10-4 millibars of diffuse helium added. Note that (b) and (c) use the same range of intensities – centred on the median intensity for each micrograph to account for the differences in raw count rate – to allow for direct comparison. The increase in the secondary beam causes an increase in noise present and reduces the available contrast...... 140 Figure 5.13 (a) Photograph of the beam entry into the pinhole plate in its original form. (b) and (c) show schematic diagrams of the beam entry to the pinhole plate before and after the alteration to improve pumping around the pinhole (respectively). The bored-out pinhole plate improves the pumping around the pinhole and results in a reduced effusive beam accompanying the main supersonic free-jet beam...... 143 Figure 5.14 SHeM micrographs of a salt crystal illustrating improvement to the pumping around the pinhole. (a) uses the original pinhole plate, and was collected using an 8 micron step size, while the pinhole plate in (b) had been opened up to improve pumping (7 micron step size). Both images were collected at otherwise identical beam and detector conditions, with colour-bar indicating the relative count rates. Scale bars are 250 microns in length...... 144 Figure 5.15 (a) Histogram of the pixel intensities for the SHeM micrographs collected of the salt crystal with the original (red) and modified (blue) pinhole plates (as seen in Figure 5.14). The difference in count rates is immediately apparent. (b) is the same data but after each set of intensities have been normalised to zero background, allowing a more direct comparison of the shape of the intensity distributions and revealing the broadening of peaks associated to certain sample features by the effusive beam...... 145

xxiv

Figure 5.16 SHeM micrographs of a honey bee (Apis Melifera) eye as imaged with different effusive contributions. (a) Replicates the larger effusive contribution of the Mark I SHeM with an additional 4.0 x 10-4 millibars of diffuse helium added to the differential chamber; (b) is the original pinhole plate; while (c) uses the modified pinhole plate with better pumping in the vicinity of the pinhole. All scans use a 2 micron step size, and all scale bars are 100 microns in length. As the size of the effusive contribution is reduced from (a) to (c), the image quality clearly increases – despite the lower total count rates...... 146 Figure 5.17 Sample mounting geometry for enhancing contrast in thin films (side and top views). The sample – in this illustration a TEM grid - is suspended over a gap in the mounting plate, allowing helium that passes through the sample to quickly become part of the background...... 149 Figure 5.18 SHeM micrographs of a section of TEM grid onto which portions of a spin-cast C60 organic film have been floated. (a) Survey scan (200 micron scale bar, 15 micron steps) of the grid showing how it has been placed over a hole in the sample slide to allow helium to pass through to form a transmission-like image. Boxed region indicates the scan area shown in (b), a higher resolution image (50 micron scale bar, 2 micron step size) showcasing the increased contrast afforded. While the transmission-like mounting immediately reveals all the holes that form in the spin-cast film, the reflection mode image contains information that would be either very difficult or impossible to determine from a purely transmissive image (for example, the placement of each piece of film on the grid surface, especially for the curled sections in the top right of the image)...... 149 Figure 5.19 Matched micrographs of a butterfly wing (Tirumala hamata). (a) Reflection optical micrograph (Leica M205 C), scale bar 600 micron. (b) Scanning helium micrograph as imaged with the Mark II SHeM with 8 micron steps, scale bar 600 micron. (c) SHeM micrograph of region indicated by square in (b) using 4 micron steps, scale bar 100 micron. (d) SHeM micrograph of another area of the wing taken with 4 micron steps, highlighting the shape of the wing scale (scale bar 50 micron)...... 151 Figure 5.20 Comparison of reflection optical (Leica M205 C) (a,c) and SHeM (b,d) micrographs of a honey bee wing (Apis mellifera) as an example of topological contrast. Bottom images taken from the square region indicated in (a). SHeM micrographs collected with (b) 8.75 micron and (d) 2 micron steps. Scale bars are 500 and 50 microns for (a, b) and (c, d) respectively...... 153 Figure 5.21 (a) SEM micrograph of a 40 nanometre thick gold University of Newcastle logo produced via electron beam lithography at the ACT node of the ANFF network. Identical logos in other metals (chromium, nickel and platinum) were produced for the purposes of an investigation into chemical contrast with scanning helium microscopy. (b) SHeM micrograph of the same sample. Image was collected with a 12 second dwell per pixel, and a 7 micron step size. Despite the clear appearance of the logo, the contrast is much smaller than any of the scans shown previously. Scale bars in both micrographs are 50 microns in length...... 155 Figure 5.22 SHeM micrographs of the 15 nanometre thick gold University of Newcastle logo partially obscured by a piece of dust. (a) Full image taken with the sample at specular, while (b) shows a section of the sample at the same position. (c,d) Small region of the sample at 500 and 1,000 μm further back from the pinhole respectively. Scale bar 50 microns...... 156 Figure 5.23 SHeM micrographs of the 40 nanometre thick University logos in different metals on pieces of the same silicon subsrate. Clockwise from top left: (a) gold (b) nickel (c) platinum and (d) chromium. Scale bar is 50 microns in length. In all four images, the intensities have been normalised relative to the silicon background in order to make a direct comparison between the metals possible. Additionally, some of the collected micrographs have been rotated, thus changing the directions so far associated with the incident beam and detector pathway...... 158

xxv Figure 5.24 SHeM micrographs demonstrating the dependence of the observed contrast for the 15 nanometre thick gold-on-silicon sample with helium mean beam energy. All scans were performed with the sample in the specular imaging position, at room temperature (294K), and using a 200 bar stagnation pressure beam. Utilising the heating/cooling system for the beam described previously, beam energies of (a) 83, (b) 72, (c) 66, (d) 42, and (e) 21 meV were achieved. Below the experimental results are simulated micrographs, testing whether count rate changes were the cause of the observed differences. Simulation 1 involved matching the average count rate and noise levels for each original micrograph, with the contrast level constant as determined from the 66 meV image. Simulation 2 went a step further, and included a variable contrast between the gold and silicon. Comparisons of the experimental and simulated micrographs led to the conclusion that there was an additional contrast mechanism at play. .. 161 Figure 5.25 SHeM micrograph of a rough silicon oxide surface which had been masked with a TEM grid and then a sub-nanometre thick gold film thermally evaporated. Even with the topographic feature provided by the carbon tape (bottom left corner), the chemical contrast allows for the location of the gold film to be seen easily (darker regions on the flat sample area). Inset shows a section of the masked surface if the count rates are limited to those due to the gold film on the silicon oxide surface. Scale bar 250 microns in length. Micrograph collected in collaboration with Dr. Sabrina Eder...... 163 Figure 5.26 Comparison of the two most likely candidates for the beam profile, namely a top hat distribution (a) and a Gaussian distribution (b). Determining the intensity profile of the beam directly is difficult – instead, by scanning the beam across a feature (typically a sharp edge) in one dimension, the line spread function (LSF) can be found...... 166 Figure 5.27 Edge Spread Function (ESF) and Line Spread Function (LSF) for the Newcastle SHeM for the most common imaging parameters. The ESF is found by scanning the beam spot across a silicon knife edge placed at the specular position (in half micron steps), while the LSF is found by numerically differentiating the ESF. Note that the ESF has had a 3 pixel moving average smoothing filter applied for reasons of noise and the ends padded to simplify the resultant LSF...... 166 Figure 5.28 Result of convolving the Gaussian fit for the LSF from Figure 5.27 with a periodic grid. The grid spacing sets the size of the dip in the central intensity for the convolution, corresponding to different resolution definitions. (a) Rayleigh Criterion (~26.5% difference in intensity), achieved by a separation of 4.8 microns. (b) A grid spacing of 2.8 microns results in the convolution obeying the Dawes Limit (~ 3% difference in intensity). The latter criteria agrees well with the observed contrast for the instrument, and so 2.8 microns is taken as the resolution for the most common imaging parameters used...... 168 Figure 5.29 Top: ESF and LSF for the Mark I SHeM, obtained from the vertical profiles shown in Figure 4.19 . The ESF has had a 3 pixel moving average filter applied and the ends padded. The Gaussian function fit to the LSF is in good agreement (R-squared value of 0.991), and yields a FWHM of (5.4 ± 0.2) microns. Bottom: Results of convolving the Gaussian beam profile for the Mark I SHeM with a periodic feature with separation of 1.5 microns to satisfy the Dawes Limit...... 169 Figure 5.30 Top: Gaussian fits to the LSFs derived from horizontal knife-edge scans for three different pinhole arrangements for the Mark II SHeM. Note that the scans have been aligned to the peak of the Gaussians for ease of comparison. Bottom: Full width half maximum values for the three different pinholes extracted from the Gaussian fits...... 171 Figure 5.31 SHeM micrographs of the edge of a butterfly wing as imaged using different pinhole diameters, namely (a) 5 microns and (b) 2 microns. Both scans were conducted with identical beam conditions, and used 1.5 micron steps (scale bars 50 microns in length). While it can be seen that the micrograph utilising the 2 micron pinhole is indeed sharper than the 5 micron, the improvement to resolution was not comparable to the reduction in pinhole diameter. Also note the effects of the reduced intensity in (b), especially the prevalence of noise...... 172

xxvi

Figure 5.32 Schematic illustrating the trajectories of atoms expanding through the nozzle of a free-jet beam source transitioning from hyperbolic arcs to straight lines known as streamlines. Tracing these streamlines back to the expansion axis, we find what is termed the virtual source point from which they appear to emanate. The virtual source plane in then defined as the plane perpendicular to the expansion axis through this point. Figure adapted from Beijerinck et. al. [101]...... 173 Figure 5.33 Schematic diagram illustrating the concept of the virtual source for a free-jet expansion. By mapping the distribution of trajectories at the quitting surface back to the virtual source plane, we may define the virtual source. The perpendicular velocity distribution of the atoms in the expansion as they leave the quitting surface is the critical parameter describing the resultant size of the virtual source. Image courtesy of Barr [32]...... 174 Figure 5.34 FWHM values extracted from the Gaussian fitting for horizontal knife-edge scans conducted at various nozzle-to-skimmer separations. All scans were conducted using a 200 bar beam with room temperature stagnation volume and a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns. No trend was observed in the data, indicating the independence of instrument resolution from the nozzle-to-skimmer separation...... 175 Figure 5.35 FWHM values extracted from the Gaussian fitting for vertical knife-edge scans conducted at different beam stagnation temperatures. All scans were conducted using a 200 bar beam and a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns. The width of the beam profile increases with increasing stagnation temperature, as would be expected considering the monochromaticity of the beam source and hence any potential broadening effects. Data is fit well (R2 = 0.999) by a linear function as indicated by the dotted line...... 177 Figure 5.36 FWHM values extracted from the Gaussian fitting for horizontal knife-edge scans performed with a range of skimmer sizes. Beam Dynamics Type 1 skimmers with nominal diameters of 100, 200 and 510 microns were used, with a nozzle-to-skimmer separation of 15 millimetres. Beam stagnation was 200 bar, room temperature, and the modified 5 micron pinhole plate formed the final optical element. A direct dependence of the beam profile width with respect to skimmer diameter can be seen, indicating the skimmer forms the restricting element in terms of resolution...... 179 Figure 5.37 Experimental LSF data and associated Gaussian fits for horizontal knife-edge scans performed with Beam Dynamics Type 1 skimmers with nominal diameters of (a) 100, (b) 200, and (c) 510 microns...... 180 Figure 5.38 Schematic of the beam geometry used for equation 5.8 describing the size of the spot produced on the sample surface by geometric optics. For the Mark II SHeM, the virtual source diameter is taken as the skimmer orifice diameter, in contrast to the original use of the formula whereby the quitting surface was used. Figure after Witham [86]...... 182 Figure 5.39 Plot of the spot diameter for the Mark II SHeM as found experimentally with knife- edge scans and via a geometric optics model for different skimmer diameters. The experimental spot size was taken as the full width tenth maximum as calculated from the Gaussian fitting to the LSFs as described previously. Both data series have been fitted with linear trends as a point of comparison...... 182

CHAPTER 6 ...... 185 Figure 6.1 The first neutral helium images as produced by Koch et. al. [58] of a hexagonal TEM grid. Micrograph (a) was produced with a beam focused down to a 3 micron spot and 8 seconds collection per pixel, while the zoomed region (b) used a 2 micron spot and 14 seconds collection time per pixel...... 186 Figure 6.2 Schematic of the optical layout for the first neutral helium images [58]. By having the helium beam pass directly through the sample to strike the detector an image may be produced in a manner analogous to transmission mode optical. The resolution of the produced TEM grid image was 1.9 ± 0.1 microns, limited by the chromatic aberrations of the zone plate acting as the focusing element...... 187

xxvii Figure 6.3 Schematic view of the instrument geometry for the neutral atom microscope. In the most recent iterations, the nozzle is held between 300 and 600 microns from the pinhole aperture while the working distance is typically between 10 and 50 microns. Together, the minimisation of the distance from source to sample enables the instrument to generate a large helium flux incident on the sample surface. Image courtesy of Witham et. al. [87]...... 189 Figure 6.4 Micrographs as produced by the Neutral Atom Microscope (‘NAM’). (a) 50nm thick crumpled gold film overlaid on mica background (scale bar 40um). (b) Multilayer graphene (scale bar 50um). (c) Crocosmia pollen grain (scale bar 30um). Images courtesy of Witham et. al. [87]...... 190 Figure 6.5 Plot of the magnitude of the topological contrast as given by equation 5.1 for a range of values of θ and δ. Note that when the condition that θ + δ < 90o is broken (ie: the detector line-of-sight is occluded - top right half of the plot), the contrast has been set to zero for readability. The region between the blue dot-dashed lines indicate the range of θ values for the Mark II SHeM, while that between the red dashed lines show the equivalent for the NAM...... 192 Figure 6.6 Plot of the signal-to-background ratio as a function of source chamber pump rate (controlled via the pump rotation speed). Signal-to-background ratios determined from reflected intensity on and off a silicon wafer. Data courtesy of Matthew Barr [32]...... 195 Figure 6.7 Plot of the signal-to-background ratio as a function of sample chamber pump rate for a section of flat silicon oxide surface. Silicon chip was mounted above a hole in a sample slide, with the background taken as the count rate obtained from the hole. Note that the experiment utilised an original Mark II pinhole plate, resulting in sub-optimal signal-to-background ratios. 195

REFERENCES ...... 203

xxviii

LIST OF TABLES

CHAPTER 1 ...... 1

CHAPTER 2 ...... 29

CHAPTER 3 ...... 57 Table 3.1 Comparison of the modelled performance characteristics of the prototype microscope with those found experimentally for a 200 bar beam at 295K. Note that the count rate, noise, and signal-to-background are all dependent on the sample under investigation – the values given are representative...... 82

CHAPTER 4 ...... 95 Table 4.1 Comparison of the distances making up the total beam path length for the two SHeM systems. The changes to the shape of the chambers mean the Mark II beam path length is much shorter, contributing to an improved count rate...... 120

CHAPTER 5 ...... 127 Table 5.1 Comparison of the quality of the micrographs presented in Figure 5.12 (b) and (c) due to the influence of the stronger effusive beam. The signal-to-background ratios and Michelson contrasts are produced by comparing the brightest and darkest features within the image, namely the silicon substrate and the TEM grid edges. It should also be noted that for the signal-to-noise ratio, the minimum intensity in both micrographs was set to zero (as discussed in the text). .. 142 Table 5.2 Image quality metrics pulled from the micrographs shown in Figure 5.16. Due to the organic nature of the sample, finding larger areas with sufficient pixels to pull viable statistics from proved impossible, hence the lack of errors for these measurements...... 146 Table 5.3 Comparison of the experimentally determined Michelson contrast for each of the metallic logos on silicon and the predicted contrast using the model of MacLaren et. al. [19]. 160 Table 5.4 FWHM values derived from Gaussian fits to vertical knife-edge scans for different beam temperatures. As a point of comparison, the approximate size of the quitting surface for each temperature (given a 200 bar stagnation pressure, and 10 micron nozzle diameter) as given by Miller [34] is included, as well as the terminal speed ratio...... 177 Table 5.5 FWHM values derived from Gaussian fits to horizontal knife-edge scans for different skimmer orifices. For the experiments in question, system employed a 200 bar beam, room temperature stagnation volume, a nozzle-to-skimmer separation of 15 mm. and the modified 5 micron pinhole plate. Resolution estimate based on the Dawes limit as previous detailed. . 179

CHAPTER 6 ...... 185

xxix

xxx

LIST OF ABBREVIATIONS

h AFM – Atomic Force Microscopy h DWF – Debye-Waller Factor h EI – Electron Ionisation h e-p – electron-phonon h ESF – Edge Spread Function h FI – Field Ionisation h FWHM – Full Width Half Maximum h FIM – Field Ion Microscope h HAS – Helium Atom Scattering h LSF – Line Spread Function h NAM – Neutral Atom Microscope h NEMI – Neutral Microscopy h PSF – Point Spread Function h RGA – Residual Gas Analyser h RMS – Root Mean Square h SEM – Scanning Electron Microscopy h SHeM – Scanning Helium Microscope h STM – Scanning-Tunnelling Microscope h STXM – Scanning Transmission X-ray Microscopy h TEM – Transmission Electron Microscopy h WD – Working Distance

xxxi

xxxii CHAPTER 1: Introduction

CHAPTER 1

INTRODUCTION

1.1. Background An iconic symbol of the natural sciences, microscopy has driven our understanding of the universe. Each new instrument has allowed us to peer further and further into realms imperceptible to the naked eye; in turn, our increasing knowledge has fuelled the development of even more sophisticated microscopes. In the modern laboratory, researchers enjoy access to an incredible array of techniques far beyond the simple arrangement of optical lenses employed centuries prior.

While photon based instruments remain the most common permutation, the last 100 years have seen the advent of the ‘matter-wave’ class of microscope. These instruments employ massive particles as the probe of the sample in an effort to bypass the fundamental limit to resolution imposed by the wavelength of light. A proposal by Louie de Broglie in 1924 [1], wave-particle duality dictates that the wavelength λ of a particle will be determined by its momentum p:

ℎ 휆 = (Eq. 1.1) 푝

(h being the Planck constant). The first microscope to take advantage of the idea was a prototype transmission electron microscope (TEM) constructed by Ernst Ruska in 1931 [2], and within a couple of years the technique offered a superior resolution to any optical system of the day. Modern electron microscopy consists primarily of the TEM and the scanning electron microscope (SEM), and can achieve resolutions below the Ångstrom level [3]. More recent additions to the matter-wave category include the scanning helium ion microscope [4] and the neutron microscope [5], but electron-based techniques remain the most common.

Whether the instrument utilises photons, electrons, or a physical tip (as employed in atomic force microscopy (AFM) and scanning tunnelling microscopy (STM)) to probe the sample, there exist a range of materials which remain a challenge to examine due to the possibility of degradation as a direct result of the imaging process [6-8]. The advent of organic and molecular electronics means that the

1 CHAPTER 1: Introduction

manipulation of soft materials (such as polymers and biological molecules) is a major new focus for nanotechnology [9, 10]. These delicate materials are very sensitive to damage from the energetic photons or charged particles used in X-ray, ion beam or electron microscopies, or the currents used in STM [10-14].

Figure 1.1 Examples of sample damage and image artefacts in high resolution microscopies. (a) The result of scanning electron microscope (SEM) imaging of a ternary blend film consisting of poly(3-hexylthiophene), phenyl-C61-butyric-acid- methyl-ester, and squaraine. All three organic components are known to suffer damage under the energetic probe, as indicated by the dark rectangles in the micrograph. Scale bar is 20um. (b) and (c) show the transmission electron microscope (TEM) and scanning transmission x-ray microscope (STXM) images (respectively) of a matched region of polymer nanoparticles on a silicon nitride window. Whilst the micrographs show that the nanoparticles maintain their structure under the electron and x-ray beams, spectroscopic STXM studies revealed their chemical structure had been permanently altered through the breaking of carbon- carbon double bonds. (d) and (e) show SEM micrographs of silver nanowires and a polyacetonitrile film respectively. Both images display the edge enhancement and charging effects that can occur during SEM imaging. Scale bars are 1um and 10um respectively. Micrographs courtesy of Natalie Holmes.

Damage can take many forms, both the obvious, and the insidious [8]. As an example of the former, Figure 1.1 (a) shows the typical electron-beam damage incurred whilst obtaining high resolution images of an organic electronic film. Regions of the sample exposed to the electron beam are easily identified by the dark rectangles. Chemical bond damage is much harder to identify. Figure 1.1 (b) and (c) show the TEM and scanning transmission x-ray microscope (STXM)

2 CHAPTER 1: Introduction

images taken of the same set of polymer nanoparticles at the University of Newcastle and at the Advanced Light Source in Berkeley, USA. The samples were prepared in Newcastle and characterised using TEM before being flown to the USA for STXM analysis. Despite the time and distance between collecting the images, it is possible to navigate to the same spot on a sample surface with a resolution of ~20 nm. However, subsequent X-ray absorption experiments revealed that while the two images demonstrate identical morphologies, the initial TEM analysis had destroyed the chemical structure of the nanoparticles, resulting in a completely adventitious carbon surface [15].

Beyond organic molecules, there exists a wide array of fragile structures which suffer degradation under traditional microscopies [6-8, 14, 16, 17]. Furthermore, the processing necessary to prepare a sample for imaging may interfere with the desired investigation. For example, samples undergoing electron microscopy are generally gold or carbon coated prior to being entered into the system – a requirement to prevent charging effects from causing image distortion (see Figure 1.1 (d) and (e) for examples), or even direct damage [8, 18]. Electron microscopy sees widespread application in biological studies, but care must be taken with appropriate coatings and accelerating voltages to produce viable images [8, 18]. Magnetic samples can play havoc with the charged probes of some techniques, and transparent thin films are not easily imaged optically [19, 20]. Considering the use many of these more difficult materials have in modern technology, developing new methods to image them is of significant import.

Figure 1.2 Comparison of the wavelength as a function of probe energy for photons, electrons, neutrons and helium atoms.

3 CHAPTER 1: Introduction

The wavelength (and hence ultimate resolution) of the photons and particles used is intrinsically linked to the energy of the probe - in other words, improving resolution requires a corresponding increase in probe energy, and thus an increase to the risk of damage. Figure 1.2 shows a plot of the probe wavelength as a function of energy for several probe species. Even for lower energy electron beams, resolutions of the order of nanometres or below require kilovolt accelerating voltages. Considering a typical atomic bond has an energy of a few electron volts, it is easy to see why these energetic probes have the potential to cause harm. One solution to the issue posed by these delicate samples lies in moving to a different probe particle entirely - neutral helium atoms.

1.2. Helium Atom Scattering Estermann and Stern first employed atoms as a probe of the structure of a sample surface in 1928 [21], recording the helium diffraction pattern produced from a LiF crystal in a confirmation of wave-particle duality for atoms. The particular choice of atom has persisted in the form of helium atom scattering (HAS); a technique which has found significant application in surface science [22]. A monochromated beam of neutral helium atoms broadly illuminates a sample, and through collection of the diffraction pattern information regarding the order of the surface may be derived.

The attractiveness of helium as the atom scattering probe of choice is due to the nature of the atom-surface interaction. As seen in Figure 1.2, a helium atom with a wavelength approaching typical crystallographic dimensions has an energy in the range of 20-100 meV; orders of magnitude lower than the more conventional photon or electron probes. The helium atoms are uncharged, chemically inert, lack spin, and have a low polarisability – consequently HAS is uniquely non-destructive, able to avoid the thermal or electronic excitations normally associated with imaging [16, 19, 23, 24]. In fact, the helium atoms refract from the outer electronic charge density contours of the surface, making the interaction unambiguously surface sensitive. Helium has a relatively large cross-section (orders of magnitude higher than electrons, photons and neutrons [19, 25]), and so HAS is additionally highly sensitive to surface defects and even very small adsorbates, including hydrogen [26]. The technique also carries with it a level of chemical sensitivity through the analysis of inelastic scattering events [19, 27]. The energy and momentum of helium atoms capable of atomic resolution are well matched to surface vibration modes and so may exchange energy through phonon creation or annihilation,

4 CHAPTER 1: Introduction

opening up the investigation of dynamic surface processes [19, 28]. Consequently, HAS is ideal for in-situ, non-invasive studies of delicate adsorbate structures [25, 28, 29], thin film growth processes [25], surface alloying [25], and phonon dispersion measurements [27, 28].

1.2.1 The Atom-Surface Interaction

The information able to be extracted via HAS derives from the nature of the helium atom-surface interaction, and as such it is worth discussing the different scattering pathways. Note that the following is a brief overview; more detailed discussions can be found in several dedicated review articles on the subject [21, 25, 26, 28, 30].

Consider a helium atom approaching the surface of an ordered solid. At distances not far from the surface, the impinging atom will feel an attraction due to Van der Waal’s dispersion forces. Should it then come too close to the surface, it will be sharply repelled via Pauli exclusion as its electronic wavefunction begins to overlap with those of the surface atoms. The steep potential barrier associated with the repulsive forces will lead to classical turning points some 2-3 Å from the surface atomic cores, with the distance further for helium atoms striking the top of a surface atom than for those falling between. The result is a locus of turning points – the actual scattering surface – forming what has been termed a ‘corrugation function’ [24, 25], following a contour of constant surface electronic charge density. Figure 1.3 shows a sectional view of a helium atom approaching a surface, including the equipotentials of the corrugation function (dotted lines).

Note that the scattering surface is a function of the atomic arrangement, as well as the electronic character of the constituent atoms. By way of an example, an insulator surface with its highly localised electron populations will present a more ridged corrugation function as compared to a conductive metal surface with its nearly free electron sea [31].

Figure 1.3 also illustrates the three trajectories available to the incident helium atom: elastic scattering (trajectory 1), inelastic scattering (trajectory 2), and resonant state trapping (trajectory 3). Elastic scattering events are more common than inelastic by several orders of magnitude [24], and form the bulk of the subsequent signal. The helium atom will backscatter to produce a diffraction

5 CHAPTER 1: Introduction

Figure 1.3 Possible scattering mechanisms for a thermal helium atom incident on a surface. Dotted lines represent equipotentials for the helium-surface interaction. Trajectory 1 – elastic scattering, yielding a diffraction pattern characteristic of the surface periodicity. Trajectory 2 – inelastic scattering as a result of energy exchange with phonons or adsorbates. Trajectory 3 – temporary trapping of the impinging atom into a resonant state of the interaction potential. Permanent trapping (adsorption) of the atom is also possible, but very unlikely for thermal helium. Figure courtesy of Barr [32], originally adapted from Toennies [33] and MacLaren [24].

pattern characteristic of the surface periodicity. Should the incident atom interact energetically with the surface, either through energy exchange with a phonon or an adsorbate, it will follow trajectory 2. The energy and momentum of a thermal helium beam is well suited to interactions with surface vibrations - indeed, the most ideal of all the noble gas atoms. The low mass and small size of the probe particle (only surpassed by hydrogen, which suffers acutely from background issues) leads to the minimum local deformation of the atomic lattice, and thus the maximum number of potential vibrational modes [30, 34]. An energy resolved analysis of the backscattered intensity (typically accomplished through time-of-flight spectroscopy) allows for a calculation of the size and strength of the exchange events [27, 28]. Resonant state trapping (trajectory 3), both temporary and permanent, is highly unlikely as compared to the first two processes [33]. The incident atom may exchange sufficient momentum with the surface to leave it momentarily trapped in the local vicinity, eventually scattering out of the resonant state through an elastic or inelastic event to join the reflected intensity. There is also the potential for the helium atom to lose sufficient energy to fully adsorb onto the surface, but the probability of such an occurrence is insignificant for thermal helium on account of its inert nature and low polarisability [24].

6 CHAPTER 1: Introduction

1.2.2 Spatially Resolving HAS

The use of helium atoms is not without its drawbacks. The very same properties which make it an ideal probe of a sample surface also mean that the atoms are difficult to manipulate, hindering efforts with regards to neutral atom focusing and detection (see Section 1.4 below). As a result, HAS is traditionally considered a signal limited technique, relying on broad illumination of the sample surface in order to ensure adequate count rates. Moving from a diffraction technique operating in momentum space to a real-space microscopy [32] is an attractive proposition – the result would be a completely surface sensitive imaging method with no risk of damage to delicate materials and the potential for atomic resolution (based on the wavelength of the probe). Such an instrument – a spatially resolved form of HAS – was first suggested in 1970 [35] and has since been termed a ‘Scanning Helium Microscope’ (SHeM). Similar to an SEM, a SHeM would selectively illuminate the surface under investigation and then build up an image by rastering the sample underneath the helium beam. The current absence of such instruments is testament to the obvious technical challenges in mustering sufficient helium intensity for useful imaging.

1.3. Contrast Mechanisms

Prior to any technological considerations for a microscope that seeks to harness neutral helium as the probe, the fundamental issue of what potential information can be revealed by the produced micrographs must be addressed. A question faced by all microscopy techniques, the answer to which is generally arrived at through the consideration of a few key factors; image resolution, noise levels, and contrast available being the most common [36]. For the proposed scanning helium microscope, resolution and noise are primarily linked to the amount of signal present, and thus tied directly to technological concerns. The question of contrast however is a function of the variations in backscattered helium intensity, and thus originates in the nature of the atom-surface interactions detailed in 1.2.1. Drawing on decades of HAS theoretical work and experiment, we can begin to investigate the sample systems that will lend themselves to SHeM imaging and evaluate the worth of such a technique to the wider scientific community.

In the context of microscopy, contrast refers to the difference in intensity or colours within a produced micrograph. Without such differences, details of the sample under investigation would not be distinguishable from the adjacent surroundings

7 CHAPTER 1: Introduction

and overall background. Furthermore, the size of the contrast variations must be greater than the noise present – a core consideration for the instrument development described in Chapter 2. Mathematical definitions of contrast are plentiful, but with regards to a scanned probe microscopy a sensible starting point is the ‘peak-to-peak’ or Michelson contrast definition routinely used to determine the quality of a signal relative to the its noise level [37]. Consider two neighbouring

pixels in a micrograph with intensities of IA and IB, the Michelson contrast C between the two pixels will then be 퐼 − 퐼 퐶 = | 퐵 퐴 |, (Eq. 1.2) 퐼퐵+ 퐼퐴 producing a contrast which will range between zero (no contrast present) and one.

In an effort to simplify and focus any discussion of potential sources of image contrast for a SHeM, here we will characterise the diverse atom-surface interactions into several broad categories: topological, chemical, and diffractive contrasts. When considering the major pathways represented in Figure 1.3, we will consider only elastic and inelastic processes, as resonant state trapping is so unlikely for neutral helium atoms as to be ignored [24]. Figure 1.4 illustrates the breakdown of the probe-sample interactions into these categories, as well as further divisions of each contrast mechanism.

Figure 1.4 Schematic illustrating the classification of the contrast mechanisms for scanning helium microscopy. The contrast available to the technique ultimately stems from the nature of the probe-sample interaction – at the highest level, an elastic or inelastic interaction. From these scattering trajectories, we may define our three contrast mechanisms: topological, diffractive, and chemical (yellow text).

8 CHAPTER 1: Introduction

Topological contrast will be the primary process for SHeM, consisting of the elastically backscattered helium atoms from a disordered surface. The ‘texture’ of the sample surface - any deviations in the mean plane originating principally from the roughness of the surface, but also factors including surface defects and adsorbate overlayers – will dictate the amount of helium directed into the specular channel. Thus, the surface morphology will influence the intensity recorded at each pixel of an image. Depending on the feature size relative to the resolution of the instrument, the morphology can be directly observed (supra-resolution) or simply influence the recorded intensity (sub-resolution). Owing to its de Broglie wavelength, room temperature helium atoms can undergo diffraction from clean, well-ordered sample surfaces. The degree of corrugation of the surface electronic potential will determine the intensity and location of the produced higher order diffraction channels. Interference effects from periodic step features such as atomic terraces will also produce similar angular variations in the scattered helium intensity. Finally, chemical contrast will originate from an energy exchange between the impinging helium atom and the surface under investigation, either through phonon or adsorbate interactions. The relative occurrence of the former is far higher, with potential energy exchanges directly with surface phonons, or those of the bulk through phonon induced surface charge-density oscillations (the ‘Quantum Sonar Effect’ [38]).

The classification shown in Figure 1.4 is, of course, a simplification, but a useful one. At a fundamental level, being able to link the physical phenomena (in this instance, the nature of the atom scattering from the sample surface) with the resulting micrograph contrast is key to all microscopy. The schematic should not be considered final, and will likely be a point of discussion for the field as more of the mechanisms are demonstrated experimentally. By means of an example, ‘chemical contrast’ as a term relating solely to inelastic scattering events is misleading. Considering the sensitivity of HAS to surface structure, defects and adsorbate layers, there is potential to be able to differentiate between materials through sub-resolution elastic scattering. Finally. it is worth mentioning that realistic scattering events are not ‘single-hit’ processes, and can in truth encompass any combination of elastic and inelastic scattering events [28].

9 CHAPTER 1: Introduction

1.3.1 Topological

Analogous to SEM [39], the dominant contrast mechanism for SHeM will be topological in nature [19]. The localised tilt of the sample surface with respect to the detector will dictate the portion of the backscattered atoms collected, and hence the surface morphology will generate image contrast. Just as in SEM, topological contrast will create micrographs with intuitive features, enhanced by the surface sensitivity of the probe particle. The majority of materials of interest will not have surfaces clean and ordered enough to generate a viable (that is, sufficiently intense [25, 30, 40]) diffraction pattern, and will instead consist of a multitude of randomly orientated and sized surface features, additionally mixed with surface defects and adsorbate layers. The elastically scattered helium from a microscopically rough surface can be generalised as following a cosine distribution, peaked normal to the mean plane of the surface [19, 24]. In reality, the relative size and distribution of surface features will dictate some ratio of both diffuse and specular (zeroth order diffraction pattern) reflections and a correspondingly more complicated angular variation in intensity. Additionally, topological contrast will include instances of occlusion, whereby either through blocking of the probe or detector there results a loss of intensity from a portion of the sample surface.

It is useful to sub-divide topological contrast a step further, relative to the resolution of the instrument. The random assortment of features below the resolution will not be individually detected in the produced micrograph. Rather, they will dictate the specular-to-diffuse ratio and hence (for a fixed detector position) the recorded intensity. Feature sizes of the order of, or larger, than the resolution will lead to localised intensity variations as expected. The need for these definitions of sub- resolution and supra-resolution scattering (respectively) is evident when considering two surface planes, both with matching relative tilts to the detector position, but different variations in feature sizes below instrument resolution. Despite no ‘direct’ sample morphology observed in a produced micrograph, the planes will have different recorded intensities as a result of elastically scattered atoms from surface features.

An expression for the 2D topological contrast from a perfectly diffusely scattering surface has been postulated in the literature by MacLaren et. al. [19]. Based on this scattering assumption, the backscattered intensity I is independent of the

10 CHAPTER 1: Introduction

incident probe direction, and depends only on the relative angle of the mean sample plane relative to the detector (θ):

퐼 = 퐼0 푐표푠(휃) (Eq. 1.3)

(where I0 is the peak reflected intensity off the surface). Consider two planes inclined by ±δ from a mean plane whose normal is itself inclined at an angle of θ with respect to the detector direction, as shown below in Figure 1.5.

Figure 1.5 Geometry for the 2D model for topological contrast from a surface with perfect diffuse elastic scattering. The diagram shows two planes (red) inclined by an angle ±δ from a mean plane whose normal (arrow) is itself inclined at an angle θ with respect to the detector direction (green).

Following the cosine distribution of the backscattered helium, relative intensities from the two faces can be found and entered into equation 1.2:

푐표푠(휃 − 훿) − 푐표푠(휃 + 훿) 퐶 = | | 푐표푠(휃 − 훿) + 푐표푠(휃 + 훿)

∴ 퐶푡표푝표푙표𝑔𝑖푐푎푙 = 푡푎푛(휃) . 푡푎푛(훿). (Eq. 1.4)

In addition to an initial expression for the topological contrast between two planes, equation 1.4 also reveals that the available contrast will increase as the mean sample plane is tilted further away from the detector. While subsequent contrast methods will prioritise specular reflections, samples exhibiting strong topological contrast may benefit from alternate scattering geometries.

1.3.2 Chemical

Although elastic scattering dwarfs inelastic scattering [24], HAS has proven to be very capable of exploiting the limited signal to study phonons and adsorbate structures [19, 22, 27, 28, 40-42]. Correspondingly, SHeM will be able to make use of the same interactions as a form of contrast. The number and energy of these

11 CHAPTER 1: Introduction

surface vibrations is known to be highly dependent on the composition and local atomic character of a sample surface [19, 31]. From HAS theory, the Debye Waller Factor (DWF) is used to describe the attenuation of the elastically scattered beam

( I / I0 ) due to the incident helium atoms coupling energetically to surface phonons [28]. One definition of the factor has the form: 2 24 푚 푇 (퐸푖 푐표푠 휙푖 + 퐷) 퐼 − 2 = 푒 푀 푘 훩퐷 (Eq. 1.5) 퐼0

and describes an atomic beam of energy Ei, mass m, incident at an angle ϕi on a

surface of atomic mass M, temperature T, and Debye temperature ΘD. D is the potential well depth of the interaction of the helium atom with the surface (~ 6 meV for a thermal helium atom) and k is the Boltzmann constant [24].

As evidenced by equation 1.5, deviations in the mass and Debye temperature of a surface will give rise to intensity variations. For example, a material with a rigid lattice (resulting in a high Debye temperature) or large molecular mass will divert less of the helium signal away from the specular channel than an amorphous (smaller Debye temperature) or low mass material. Thus the backscattered intensity is capable of imparting information about the composition of the surface. Furthermore, the dependence on Debye temperature enables contrast even in heterogeneous samples with thermal or structural differences, providing exciting imaging possibilities.

Consider two ideal, smooth, crystalline materials with differing atomic masses (MA

and MB) and surface Debye temperatures (ΘA and ΘB). Combining equations 1.2 and 1.5 produces an expression for the Michelson contrast due to chemical effects [19]:

훼 1 1 퐶 = 푡푎푛ℎ { ( − )} , (Eq. 1.6) 2 푀퐴훩퐴 푀퐵훩퐵

where 2 24 푚 푇 (퐸푖 푐표푠 휙푖 + 퐷) 훼 = . (Eq. 1.7) 푘

MacLaren et. al. [19] predicted that for a gold and silicon material pairing, contrast values as high as 0.5 were possible, validating the potential for SHeM contrast based on sample composition (and providing a useful first choice of sample system).

12 CHAPTER 1: Introduction

The energy of the incident helium atom dictates the dynamic surface processes it is able to interact with, and recent work has shown that the options are not restricted merely to surface phonons. Low energy (thermal) collisions will indeed interact strongly with these vibrations, but it has been found that higher energies can permit coupling to electronic charge-density oscillations [38, 43, 44], opening up additional inelastic scattering pathways. Of particular interest to the field of helium atom scattering is the electron-phonon (‘e-p’) interaction, whereby vibrations of the bulk atomic lattice can induce fluctuations of the electron population [38]. The e-p interaction has a relatively long range, allowing subsurface vibrations to couple to electron charge-density oscillations at the surface [38]. Restated, the bulk properties of a sample may impact the surface electron corrugation and hence cause inelastic scattering of helium atoms additional to that of surface phonons.

Phonon induced surface-charge oscillations have been termed the ‘Quantum Sonar Effect’ [38], and potentially play a more significant role in inelastic scattering than previously understood. The extent to which helium atoms can perceive atomic displacements several layers beneath a surface is an active area of research, one which is helping to address issues with DWF formulations based around strictly surface processes. Definitions such as that of equation 1.5 will produce excellent agreements with experimental results for some material systems, and yet for others be wholly inadequate [45]. Efforts are currently focused on redefining the DWF in terms of the e-p coupling strength [46], an avenue of research that will greatly benefit the pursuit of quantitative expressions for SHeM contrast.

A final note on inelastic scattering events is the potential for energy exchanges with surface adsorbates. Due to the helium atoms not having any internal degrees of freedom (unlike, for instance, molecules) to complicate the projectile-surface interaction, HAS has been effectively used to study the vibrational structure of adsorbates [25-29]. As a prospective contrast mechanism however, their use may be severely limited due to the available helium intensity. With elastic collisions dominating over inelastic, and phonon based interactions far more prevalent than adsorbate ones, there may simply not be enough signal to generate useful images. Should technology regarding neutral helium focusing or detection see significant improvements (see Section 1.4), the option exists to use the scattering pathway as a measure of local surface coverage, desorption events, or perhaps even tailored interfaces between adsorbates and novel substrates [47].

13 CHAPTER 1: Introduction

1.3.3 Diffractive

Samples amenable to diffractive contrast bear the greatest resemblance to those commonly studied in traditional HAS experiments. Thermal helium atoms have a wavelength comparable to typical crystallographic dimensions and so may undergo diffraction from clean, well-ordered crystalline surfaces [25, 30]. The helium atoms refract in the outer electronic potential of the surface, the shape of which is indicative of the nature of material under investigation (conductor, semiconductor, or insulator) [19, 31]. Insulators and semi-conductors will have a more localised electron distribution, causing the contours of the electronic density to be significantly more corrugated than those originating in the nearly free electron sea associated with conductors. Thus, for an idealised sample surface if one were to monitor the specular channel, the intensity from a clean metal surface should be greater than that from a semi-conductor such as silicon [19, 31]. Alternatively, positioning the detector over a particular diffraction peak off-specular would yield a SHeM imaging mode equivalent to dark-field imaging in the TEM or optical microscopes [31]. Observation of the changing intensity would reveal details of symmetry or thin film growth (island formation) within the area of the sample surface exposed to the incident helium beam (as opposed to the broad illumination of HAS) [25].

The wavelength of the helium atoms also leads to a sensitivity to the vertical morphology of atomically stepped surfaces through interference scattering from adjacent terraces [25, 30]. Furthermore, the presence of step edges or other defect sites in a surface (including adsorbates) will perturb the electron density contours such that the impinging helium atoms scatter away from the specular channel. As such, the surface reflectivity is inversely proportional to the surface defect / adatom coverage [24, 25]. Results from HAS studies indicate that concentrations of adsorbates can be detected in concentrations less than 0.1% of a monolayer [26], and one would then accordingly expect SHeM to excel at monitoring processes such as thin-film growth.

14 CHAPTER 1: Introduction

Figure 1.6 Schematic drawing of isolated (low-coverage) CO molecules on a Pt(111) surface. The solid line represents the corrugation function (‘repulsive wall’) for a room temperature helium beam. The dashed and dotted semi-circles

correspond to the cross-sections of adsorbed CO (ΣCO) and gas phase CO (σCO) respectively. The large cross-sections for helium scattering from these adsorbates are also given, illustrating the sensitivity of a helium beam to individual adatoms. Figure from [25].

It is important to note that in order to derive information from the produced diffraction patterns, modern HAS systems require a very high angular resolution (typically better than half a degree, with more advanced systems being able to resolve to down to 0.1o [25, 48]). Sensitivity to diffractive and interference effects in a SHeM would likely require similar resolution, else the detector aperture would simply collect the closely spaced diffraction peaks. The benefits associated with such an imaging mode would be well worth the long collection times resulting from a reduced intensity. Diffractive contrast tells us of the local order of a sample surface, even if there is no topological or chemical differences present. The capability to spatially map crystallinity would be a boon to a variety of fields. One example is the area of organic photovoltaics, where the crystallinity at interfaces within the layered devices has been shown to directly affect the final efficiency [49]. Further extensions of the SHeM to subjects conventionally suited to HAS, including thin film growth, adsorbate coverage, structural phase transitions, alloying, and many more, make for a compelling argument as to the benefit of such a technique. The SHeM would be sensitive to features of an atomic scale, even if the lateral resolution of the instrument (as dictated by the spot projected on the sample surface) is not.

15 CHAPTER 1: Introduction

1.4. Elements of a Neutral Atom Microscope With the unique properties of neutral helium as an atomic probe of a surface known, focus then turns to the challenge of actually constructing a SHeM. Like any other microscope, a SHeM can be thought of consisting of a few key components:

 A source of the probe particle – in this case, an intense, monochromatic beam of neutral helium atoms.  Some form of optical element capable of isolating the helium beam to a specific region of the sample surface.  A detector with which to convert the backscattered helium atoms into an electrical signal.

The desire to work with neutral helium as the probe particle leads to significant technical challenges with these components – especially the focusing and detection - and so some discussion of each is warranted.

1.4.1 Atomic and Molecular Beam Sources

As a logical extension of HAS, SHeM shares many of the same elements - most notably the beam source. Atomic beams form an integral part of many fields, with a considerable percentage of all beams produced comprised of helium. Due to this popularity, helium beams are a well-studied and modelled phenomena [30, 34, 50]. We can broadly group the different types of beams into two classes: effusive and free-jet.

The concept of 'Knudsen flow' describes a region where the characteristic dimension of the flow space is smaller than that of the mean free path of the atoms within [34]. In describing such an occurrence, it is useful to define the Knudsen number (K), namely 휆 퐾 = , (Eq. 1.8) 퐿

where λ is the mean free path of a particle, and L is a length representative of the flow space [51]. For Knudsen flow to dominate, Knudsen numbers >> 1 are typically quoted. In this environment, the atoms interact with the edges of the 'Knudsen Cell' more often than with each other. Introducing an aperture into the side of the cell then causes an effusive beam to be produced. An effusive beam source thus utilises such a cell, often incorporating a heating element to produce a suitable vapour pressure of the desired atomic species [51, 52].

16 CHAPTER 1: Introduction

A point of interest in later discussions (see Section 3.2), the shape of the effusive beam produced by a Knudsen cell will depend on not only the Knudsen number, but also potentially a dimensionless parameter β which describes the nature of the channel through which the beam expands [34, 51]. For a channel of circular cross- section (Figure 1.7), β can be defined in terms of the radius (r) and length (L) of the channel:

2 푟

훽 = . (Eq. 1.9) 퐿

Figure 1.7 The shape of the effusive beam caused by molecular flow through a channel of length L and radius r can be determined by considering the Knudsen number K and the parameter β.

A wide variety of angular distributions are possible, ranging from the familiar cosine distribution most commonly associated with these beam sources, to sharp distributions centred around the central axis of the orifice [34, 51, 53]. While still employed in many deposition applications, within surface studies effusive beams have generally been supplanted by the more modern free-jet beams due to the increased intensity and lower energy spread. For an effusive beam source, the mean free path of the atoms within the cell must fall below the size of the aperture, thus imposing a limit on the cell pressure (and consequently beam intensity) which may be obtained. Intensities may be increased somewhat by heating the atomic species further or increasing the nozzle diameter [34], but the requirement for Knudsen flow restricts the produced intensities below that required for a practical helium atom scattering system (and hence microscopy applications). Additionally, due to the large range of particle velocities as they leave the cell, the energy spread of effusive beams is also larger than is desired for such applications [51].

The supersonic free-jet beam source was introduced in 1951 [54] and is based around the expansion of a high pressure gas through a small hole into a vacuum chamber (Figure 1.8). Firstly, the atomic species is pumped to a high pressure within a reservoir behind a small aperture (typical diameters of a few microns), ensuring a Knudsen number << 1. The gas is then allowed to expand adiabatically

17 CHAPTER 1: Introduction

into a vacuum, with frequent collisions during the expansion narrowing the velocity distribution of the atoms. The resulting beam achieves sonic and then supersonic speeds in quick succession, with the expansion dropping the density until free molecular flow conditions are reached (a point termed the ‘quitting surface’). Past the quitting surface is a region of free-molecular flow (the ‘zone of silence’) where the atoms follow desirable straight-line trajectories. The shock waves at the edges of the expansion eventually recompress the gas to mark what is termed the Mach disc. With the gas atoms becoming sub-sonic at this point, the central portion of the beam is typically sampled from the zone of silence via the insertion of a cone- shaped aperture (skimmer), producing a collimated beam of atoms [34].

Figure 1.8 Schematic representation of a supersonic free-jet expansion. A dense volume of gas is allowed to expand through a small nozzle into a region of lower pressure. The high numbers of interatomic collisions in the initial turbulent flow gives rise to a supersonic expansion outwards from the nozzle. The rapidly decreasing density leads to the flow regime transitioning from fluid flow to free-molecular, with the boundary between traditionally termed the ‘quitting surface’. Past this point in the expansion (the ’zone of silence’), the gas particles are considered to have a large mean free path and travel along straight-line trajectories. It is typical to sample the centre of the expansion through the insertion of a sharp conical aperture or skimmer into the zone of silence. Schematic adapted from [24].

18 CHAPTER 1: Introduction

Supersonic free-jet sources are much higher in intensity than effusive sources, with recent powerful systems reporting numbers of the order of 2 x 1020 atoms per second per steradian [55] through optimisation of the relationships between the stagnation volume pressure, nozzle diameter, and the pump rate on the expansion chamber. Due to the nature of the expansion, the atoms in the produced beam will have a sharply peaked distribution about the mean perpendicular velocity (that is, along the beam axis) resulting in a highly monochromatic beam [56]. By adjusting the temperature of the nozzle, the energy and thus wavelength of the probe particles is able to be finely controlled, meaning that free-jet beams sources are ideally suited to microscopy applications.

1.4.2 Neutral Atom Optics

As with all microscopies, in order to image using neutral atoms as the probe particle we need to produce a controlled region of illumination on the sample surface. Ideally, we need to employ optics in order to control resolution (through the size of this region) and increase the amount of useful intensity. We can divide up the potential focusing mechanisms for neutral particles into the general categories of diffraction, reflection and refraction (Figure 1.9).

Figure 1.9 A summary of the potential methods for focusing neutral atoms to a point (Image courtesy of [32], adapted from MacLaren [24]).

Neutral atoms are already difficult to focus but helium atoms are especially problematic, as in addition to the lack of charge they have a low polarisability and no spin [24]. For these reasons, refraction based focusing of neutral helium is not possible, leaving diffraction and reflection available. Of the remaining options, the two techniques to have received the most attention are Fresnel zone plates and single crystal mirrors, both of which have been successfully harnessed to yield

19 CHAPTER 1: Introduction

focused spots of helium. Fresnel zone plates are comprised of a concentric grating of interspaced open and opaque circular annular rings through which the wave- functions of incident particles may diffract (see Figure 1.10 below). The size and placement of the rings lead to constructive interference and the generation of a focused spot [23, 57]. The fabrication of a zone plate for particles is a significant technical challenge, as the width of the smallest zone will determine the resolution of the subsequent focusing [23, 57, 58].

Papers by Doak et. al. [23] in 1999 and Koch et. al. [58] in 2008 both demonstrate the principle for helium, with zone plates with outer zone widths of 100 and 50 nm respectively, reporting central spot sizes of the order of two microns. In particular, the latter paper uses the optic element to generate an image of a grating using neutral helium in a manner similar to a rudimentary TEM. However, the collection time for this experiment demonstrates the second problem with the zone plates: intensity. The images produced took over 14 hours to complete due to the low levels of signal entering the detector [58]. The nature of the focusing is such that only 10% of the flux incident on the zone plate can be focused into the central spot and is overlaid on a broad background signal, lowering the signal-to-background ratio. As intensity is a critical consideration for the viability of the SHeM, the presently low count rate represents a substantial challenge for the effective use of zone plates.

Mirrors for helium may also be created to focus the particles through reflection [59], but the production of these mirrors is potentially even more difficult than that for a zone plate. Typical surfaces will specularly reflect much less than 1% of the helium incident on them [24], and so efforts have been directed at increasing this value. The basis of a neutral atom mirror is an ultraclean, atomically flat surface; normally either a quartz substrate ground into a parabolic shape, or an electrostatically bent single crystal of hydrogen-passivated silicon [60], although thin metal crystals have been recently shown to also be viable [61]. Reflection may be improved by the deposition of a thin overlayer which flattens the electronic corrugation with which the helium atoms will interact as much as possible. Multiple variations on the idea have been trialled, with some designs reporting helium reflectivities of approximately 20% [24, 59, 61-65]. Such mirrors can focus a large number of helium atoms to the desired spot, avoiding the intensity limitations of a zone plate, and can manifest micron spot sizes [24, 60-62, 64]. Despite this, the difficulties associated with their fabrication, transport and operation means their current state of development has fallen behind that of zone plates.

20 CHAPTER 1: Introduction

Figure 1.10 SEM micrograph of a section of a free-standing silicon nitride Fresnel zone plate used to focus helium beams. The inset shows the outermost zone with an approximate periodicity of 96 nm, demonstrating the technical skill necessary to create structures capable of interacting as desired with neutral helium atoms. Image courtesy of Reisinger et. al. [58].

Based on the work with both zone plates and neutral atom mirrors, the technology required to focus helium does exist, but the optics themselves are not easy to produce and require further research before they are suitable for implementation in a prototype microscope. The improvements in our ability to fabricate materials at very small length scales with the maturation of nanotechnology should ensure the optics will readily improve in the near future. In the interim, one potential option is the use of micron sized pinholes to produce a spot on the sample surface through collimation of a helium beam, instead of focusing. If the intensity of helium atoms from a supersonic free-jet source is high enough, then employing a pinhole to reduce the spot size and form an image is feasible, given a sensitive enough detector (see Chapter 2).

21 CHAPTER 1: Introduction

1.4.3 Detection

The final necessary component for a neutral helium microscope is a detector capable of transforming the flux of backscattered helium particles into a measurable electronic signal. Neutral detection, in even its most basic form, remains a significant challenge for the field of atomic beams, and working with helium again compounds the issue (the same properties which make it an ideal probe particle also render it among the most difficult to detect). With regards to application for a microscope, an ideal detector would need to be highly sensitive, have good spatial and temporal resolution, and be compact enough to facilitate a desktop apparatus.

There are a wide range of gas sensor technologies available which are capable of neutral detection, but none of available methods meet all of the aforementioned criteria for a microscope [66, 67]. In general, we may class gas sensors as either chemical or physical, with the former being more common except in advanced applications such as chromatography and mass spectroscopy [68]. Chemical sensors are usually adsorption based, using materials with large surface areas (including metal oxides, porous silicon, polymers and carbon nanotubes) to form an active layer. The adsorption of a molecule or atom on the surface changes its electrical resistivity, providing the means for detection. The nature of the detection means that these sensors are best suited to molecules which will easily bond to surfaces, as opposed to inert species such as helium. Correspondingly, we turn to the physical class of sensors which predominantly revolve around the ionisation of the species of interest and subsequent signal generation through traditional charged particle detectors, or the fingerprinting of the unique breakdown voltages [69, 70]. Within this group of detectors there are many variations, primarily based on the method of ionisation, with the sensitivity of the instrument directly linked to the efficiency of that method.

The most common means of ionising the neutral species of interest is that of electron impact ionisation, now more commonly referred to as electron ionisation (EI). Any neutrals which enter the detector are exposed to a flux of energetic electrons (typically around 70 eV) which induce the removal of the electrons of the atom [69, 71]. The passage of the electron near the particle distorts its electric field, and if the frequency of the electron corresponds to that of an electronic transition in the particle then an energy transfer may occur. 70 eV is chosen as a good balance between the impact cross-section and supplying enough energy to remove

22 CHAPTER 1: Introduction

an electron entirely (leaving a positive ion) for the majority of species [72]. The new ion is then able to be accelerated towards a channel electron multiplier (CEM) or Faraday cup, and hence generate an electrical signal.

While the process is sufficient to ionise most organic compounds without trouble, the low cross-section for impact between an electron and a helium atom, as well as helium having the highest ionisation energy of any species (24.6 eV), limit its effectiveness. Consequently the ionisation efficiencies for standard electron impact detectors for helium generally range from 10-6 to 10-4 - restrictively low for the application in a SHeM [66]. Recent improvements work by magnetically confining the electrons to a larger region, thereby increasing the ionisation volume and hence the efficiency of the ionisation. Efficiencies of up to 7 x 10-3 have been reported from these detectors [73, 74]†, but inevitably there is a significant degradation to the temporal response of the detector due to the sheer size of the ionisation volume (see Figure 1.11). As the spatial resolution possible with the detector is directly related to the size of the ionisation volume, and long imaging times removes the ability to investigate dynamic surface processes, work is still required in order to develop EI detectors in this style to a point suitable for microscopy applications.

Figure 1.11 The Cambridge Spin-Echo Neutral Helium Detector, demonstrating the necessary increase in size to compensate for the low efficiency of the electron ionisation process. Image courtesy of David Chisnall.

Besides EI detectors, other commonly used methods include photon ionisation and surface ionisation (with the latter sometimes referred to as a Langmuir-Taylor detector). Photon ionisation uses high energy photons to bombard the species of interest in the gas phase, typically through high powered ultraviolet lasers [75]. The

† Resulting in sensitivities of approximately 0.38 A/mbar for helium

23 CHAPTER 1: Introduction

primary process is multi-photon ionisation, whereby an electron will absorb multiple photons allowing it to move to progressively higher orbitals until it is removed, thus creating an ion [75, 76]. While the ionisation probability is close to 100% even for helium, the technique is limited by the number of photons the lasers can produce over an extended period of time. Pulse rates are low enough to restrict the overall efficiency to the order of 10-10 [67]. Surface ionisation on the other hand often has a very large efficiency, but suffers instead from selectivity issues. If a neutral atom lands on a hot (often metallic) surface, then electron interchange due to thermally supplied energy may form a new ion as it is ejected through thermal desorption [77, 78]. However, for this process to occur the energy of the neutral species must be smaller than the work function of the hot surface. The high work function of neutral helium means that it is unable to be ionised in this fashion.

There has been an effort in recent years to produce a neutral detector based around the process of field ionisation (FI), a quantum mechanical process first predicted by Oppenheimer in 1928 [79]. Under the influence of a large electric field (of the order of volts per Ångström), an electron can be enabled to tunnel out of a neutral atom to leave behind an ion [80]. In its most common form, FI is achieved much the same way as was used by Muller in the field ion microscope (FIM) [81, 82]. A sharp metallic tip, commonly made from electrochemically etched tungsten as seen in STM, is biased with a positive voltage to create a large electric field near its apex. If a neutral atom comes within a certain critical distance of the tip, the potential well of an electron within the neutral species can be distorted enough to produce a potential barrier through which the electron may tunnel.

Figure 1.12 SEM image of a typical electrochemically etched tungsten tip with end radius of the order of 30 nm.

24 CHAPTER 1: Introduction

Field ionisation offers many potential benefits for application within a gas sensor: high ionisation efficiencies, excellent spatial and temporal resolution and a compact size. It has already been employed in a small number of gas sensors due to it being a 'soft' ionisation technique [66, 80] (produces very little particle fragmentation). Soft ionisation is particularly important for large organic molecules where fragmentation increases the difficulty of subsequent mass analysis. It must be noted however that the probability of tunnelling is strongly dependent on the field strength, and so small distances away from the tip apex the ionisation probability rapidly drops off. As such, the ionisation volume for a single tip is small, leading to FIs main disadvantage: low sensitivity. The small ionisation volume means that only a small ion current is produced; a limitation that has so far prevented the technique from becoming commercially viable. Work has been started with regards to moving from a single tip to an array of field ionising elements to combat the restrictive ionisation volume [68, 80, 83-85], but it has yet to reach a point whereby it can be implemented into a microscope.

While there are many promising technologies on the horizon, it can be said that the field of atomic beams has yet to develop a universal detector which is both as fast and as efficient as would be desired for microscopy applications [66]. Until these technologies come of age, focus thus shifts to maximising the available signal in instruments in other ways – a critical consideration for the pinhole helium microscopes detailed within this thesis.

1.5. Neutral Helium Imaging

The first step towards realising an instrument capable of harnessing neutral helium atoms as a probe was conducted in 2007 by Koch et. al. [58]. By adapting an existing molecular beam apparatus through the addition of a zone plate, they were able to focus the beam and allow for spatially resolved imaging. The detection stage sits directly behind the sample (relative to the incident beam) meaning that the images are formed in a manner analogous to transmission optical or electron microscopy. Figure 1.13 below shows images collected of a hexagonal TEM grid.

During the course of the work presented in this thesis, another research group has developed an instrument geometry capable of producing reflection mode images with a resolution better than half a micron [86] by challenging some of the long- standing conventions of supersonic beams. Termed ‘neutral atom microscopy’

25 CHAPTER 1: Introduction

Figure 1.13 The first neutral helium images as produced by Koch et. al. [58] of a hexagonal TEM grid. Micrograph (a) was produced with a beam focused down to a 3 micron spot and 8 seconds collection per pixel, while the zoomed region (b) used a 2 micron spot and 14 seconds collection time per pixel.

(‘NAM’), the core departure from a more traditional geometry is the marrying of the skimmer and a downstream pinhole into a single aperture, allowing the source and sample to be brought as close as possible. A schematic of the instrument geometry can be found in Figure 1.14, showing the ‘conical aperture holder’ in the position normally reserved for the skimmer, but inverted in direction along the optical axis. The sample is rastered underneath the beam to produce an image, with the detector monitoring the helium population through a conical aperture placed off to the side. The count rates are sufficiently high that the pinhole diameter can be brought down to as low as a few hundred nanometres, affording the technique its impressive resolution.

Figure 1.14 Schematic view of the instrument geometry for the neutral atom microscope. In the most recent iterations, the nozzle is held between 300 and 600 microns from the pinhole aperture, while the working distance is typically between 10 and 50 microns. Together, the minimisation of the distance from source to sample enables the instrument to generate a large helium flux incident on the sample surface. Image courtesy of [87].

26 CHAPTER 1: Introduction

1.6. Thesis Overview As detailed in this chapter, the beneficial properties of neutral helium as a probe particle have long been a subject of interest in the literature. The aspiration to take further advantage of helium atom scattering and construct a spatially resolved microscopy had thus far been precluded by the technical limitations of the various elements which together would make up such an instrument. The original research presented in this thesis describes the design, construction, and testing of an initial prototype scanning helium microscope (SHeM) design, as well that of a subsequent instrument which iterates on its predecessor to further push the field forward.

The work is arranged as follows. Chapter 2 explains the design considerations and challenges involved in attempting to build a practical microscope with the available technology at the time. The resulting prototype instrument (the ‘Mark I’) was constructed at the Cavendish laboratories, Cambridge, and the results from preliminary testing are covered in Chapter 3. With the knowledge earned from the work performed with the prototype system, attention was then focused on building a second iteration at Newcastle. Chapter 4 contains a detailed rundown of the design and performance improvements of the ‘Mark II’ instrument, while Chapter 5 provides an in-depth examination of the new research it made possible. Of particular note are studies concerning image formation, secondary beam effects, the available contrast modes, and an initial investigation of instrument resolution. Finally, Chapter 6 concludes by taking a short look at the current state of the field of neutral helium microscopy and where we are likely to see advances in the years ahead.

27 CHAPTER 1: Introduction

28 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

CHAPTER 2

THE MARK I PROTOYPE SHEM – DESIGN OVERVIEW

The elements required to piece together a scanning helium microscope, and indeed the new physics such an instrument had the potential to investigate, have been well known for several years. However, the prevailing wisdom was that some of those elements were not yet advanced to the stage that a functional instrument could be built; specifically both the focusing and detection of neutral atoms. An opportunity to work with the people responsible for much of the early investigations into neutral helium microscopy arose, and with it an interesting question: would it be possible to build a viable instrument with the current level of technology at our disposal? An in-depth analysis of the requirements would yield useful information on the subject no matter the eventual outcome, and so work was begun on the design of a prototype SHeM. The modelling conducted in pursuit of such a design, including both the helium signal available for imaging as well as the physical layout of the various vacuum chambers, is detailed here. When the predictions of the modelling proved positive, the prototype (designated as the ‘Mark I’) was constructed at the Cavendish laboratories at the University of Cambridge.

The chapter that follows discusses each of the major components of this first attempt at a functional scanning helium microscope; with experimental results appearing in the following chapter. The instrument was a collaboration between researchers at both the University of Newcastle and the University of Cambridge, with some of the work performed also appearing in the thesis of Matthew Barr [32]. The author focused on the design work for the system, completed with the aid of Dr. Andrew Jardine, and assisted Barr with the gas flow modelling undertaken concurrently. Construction at the Cavendish laboratories was completed with the assistance of Barr and Dr. Paul Dastoor.

29 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.1. Development Process

2.1.1. Initial Investigations

Even though the general idea behind a microscope capable of utilising neutral atoms, and in particular helium, to image surfaces has appeared in the literature for many years [19, 31, 35], as of 2011 one had not been constructed; illustrating the inherent difficulties in working with neutral species. As a point of contrast, the helium ion microscope had been developed to the point of commercialisation by 2007 by Carl Zeiss Inc. [88] due in part to being able to adapt technology such as that used in scanning electron microscopy for manipulating charged particles. With this in mind, it was entirely possible that without the refinements discussed in Chapter 1 the microscope might not be feasible. In order to avoid wasting time and resources, the helium signal for the instrument was to be modelled to determine if an image of acceptable quality would be possible. With regards to the quality, it was proposed that such a criterion be based on: (a) the helium count rate, (b) the contrast between two different materials, and (c) the number of pixels in an image. The latter may also be considered a measure of the scan time, once a dwell time per pixel is established. For an initial investigation, a simple simulation was produced by Dr Andrew Jardine which yielded an image of two different materials with the ability to change the aforementioned parameters. Noise in the simulation was modelled as shot noise based on the count rate [89]. By generating sets of images for contrasts of 2%, 10% and 30%, and count rates from 10 to 1x105 counts per second, some initial boundaries on quality could be established. A minimum image size was the first parameter to be locked down to 64x64 pixels and above, which (assuming a one second dwell time per pixel) correlates to a scan time of 1.1 hours. As can be seen in Figure 2.1 below, increasing the pixel count produces clearer images of the two regions despite keeping the same contrast and count rate [89, 90].

Figure 2.1 Sample simulated micrographs for two materials with 10% contrast and a count rate of 1000 Hz. From left to right, the number of pixels in the sample images are as follows: 16x16, 32x32, 64x64, 128x128, and 256x256. As the pixel count increases the proportional effect of shot noise is diminished, resulting in a clearer distinction between the materials.

30 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

With the pixel count restricted, the minimum image quality could be further fine- tuned. As 10% topological contrast was deemed a sensible lower estimate based on the existing contrast work (based on a cosine scattering from the surface [91], it was calculated that a 5 degree difference between two surface planes would yield a 10% difference in count rates), a count rate of 1000 Hz should be sufficient. Stronger or weaker contrasts would of course bring the required count rate down or up – see the dotted line in Figure 2.2 for a visual idea of the equivalent scenarios. Such calculations were to provide only a general guidepost to the feasibility of the microscope design, as extending the collection time per pixel (as opposed to increasing the pixel count) would also work to improve image quality, albeit to a much lesser extent.

Figure 2.2 Sample 64 x 64 pixel images produced for a range of count rates and material contrasts. In the planning process for the prototype, it was decided that the build would go ahead if an image of a quality greater than a set level could be produced, represented by the images to the right of the dotted line in the above figure.

The next stage was to model the count rates available to a prototype SHeM. As preliminary designs for the various vacuum chambers and components which make up the microscope were being drawn up in CAD programs, an Excel model of the helium path through the instrument was also built using traditional gas flow equations [92]. With both of these processes running in tandem, the design evolved as an iterative process - changes to the physical geometry were fed into the Excel model, which then suggested further changes to the geometry. The design process also had benefits beyond allowing for a decision as to whether to proceed; it allowed refinement of the design before any fabrication had begun, giving the project the greatest chances for success.

31 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.1.2. HAS Geometry Comparisons

The starting point for the design of the prototype is the same as that of the technique itself: a helium atom scattering system. Typically, a HAS system consists of a source chamber, multiple differential pumping stages incorporating a chopper, a sample chamber, further differential pumping stages, and finally a detection stage. For instance, below is a schematic of a HAS system located at Bell Laboratories [41].

Figure 2.3 A traditional helium atom scattering system; in particular the apparatus at Bell Labs as described by R. B. Doak in [41]. Designs such as this were the starting point for the layout of the prototype SHeM to be built at the Cavendish laboratories.

Altering the geometry to spatially resolve the technique requires a number of changes. Given the size of the beam source and related infrastructure, in a HAS system the sample stage typically rotates to move the sample through the range of incident beam angles required. Then, through use of bellows and several mounting points on the sample chamber (as can be seen in the above diagram), a large range of detector angles can also be accomplished. In the SHeM, the incident beam and detection angles would have to be fixed at 45o to ensure there was enough space for the necessary components and the potential for strong topological contrast. The sample stage would still be controlled, but it would instead need to be able to raster a sample underneath the stationary beam to build up an image.

The second major change is the introduction of some form of optics prior to the sample to ensure only a small section of the sample is under illumination at any one time. As the spot size sets the resolution of the instrument, and heavily impacts

32 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

the helium count rates at the detector, it is a critical design consideration. HAS systems typically broadly illuminate the sample under investigation due to the samples themselves usually consisting of sections of single crystal material, and the ubiquitous desire for higher count rates. It was found through prior modelling that placing the optics between the source and sample was preferable in terms of the count rates to placing them between sample and detector. The choice of what form the optics will take then further specifies positioning. At the time of construction, focusing methods for neutral atoms were not yet in a form feasible for implementation into a working microscope. In the absence of being able to focus a greater percentage of the beam down onto the sample, the instrument would have to settle for some other means of selective illumination – namely a simple pinhole, thus making the prototype a pinhole microscope (analogous to a pinhole camera). The nature and position of the pinhole will be discussed later in the broader context of the microscope geometry.

Traditional HAS systems have to balance the requirement for good differential pumping a longer beam path affords against the reduced count rates [30, 40, 41]. The former normally takes precedence, resulting in HAS systems typically being very large (see scale bar on Figure 2.3). The limitation of implementing a pinhole in terms of available signal means making differential pumping the priority is not necessarily the case for a scanning helium microscope. Assuming the cross- sectional area of the detector to be large compared to the effective source size (a safe assumption for a stagnation detector), the intensity of the incident beam will decrease as the square of the source-to-detector distance. A shorter beam path length can thus be used to compensate for low count rates, at the cost of signal- to-background ratio. Resolution also extends beyond simply the spot size as set by the optics – the angular resolution of the detector (set by the aperture from the sample chamber into the detector volume) is also a consideration, especially to diffractive effects. The net result for the prototype design was that in order to produce a working instrument with the current equipment, the beam path length must be as small as possible to optimise the count rate.

33 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

The available signal will also directly depend on the quality of the detector used for the neutral helium. While the sensitivity of commercial quadrupoles for neutral helium can vary by several orders of magnitude [66], some companies offer versions optimised for light species. From prior work with helium beams, the Cavendish laboratories were already in possession of a HAL/3F-PIC residual gas analyser from Hiden Analytical with a recorded sensitivity for neutral helium of 3x10-5 A/mbar [93]. While among the highest for a commercially available quadrupole, the sensitivity was still restrictive - to compensate, the detector would be operated in stagnation mode.

As a final note on the influences to the instrument design, certain parts from a prior vacuum system at the Cavendish laboratories designed to test atom mirrors [24, 60] were to be repurposed for the prototype. In particular, the source chamber, the frame upon which it rests, and the infrastructure already set up (pumps, gas lines, etc.) set certain aspects of the design. Details on these recycled parts are given as required below.

2.1.3. Instrument Overview

In the interest of simplifying the explanation of both the physical elements of the microscope and the Excel model, a brief overview of the final design of the prototype is useful. Figure 2.4 shows a schematic of the instrument in its final configuration. A supersonic free-jet expansion is created in the source chamber before the centreline of expansion is selected out via a sharp conical skimmer to pass onto a differential pumping stage. The beam is directed towards the pinhole, which is mounted in a plate attached to the inner wall of the sample chamber. The result is a thin beam of helium striking the sample, mounted on two piezo stages allowing it to be rastered back and forth under the beam. Any excess helium from the differential stage is pumped away. The reflected helium from the sample is allowed to pass through to the detector via a second aperture in the pinhole plate, where the helium is allowed to stagnate and a reading of the helium partial pressure made using the Hiden quadrupole. By rastering the sample back and forth and registering the helium signal at each point, a picture of the sample can be generated.

34 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.4 Schematic diagram of the prototype SHeM. The helium beam source consists of a free-jet expansion, of which the centreline is selected out with a skimmer in the source chamber. The beam passes through a differential pumping stage to the pinhole optics of the instrument, a 5 micron FIB milled pinhole. The result is a thin beam of helium striking the sample surface in the sample chamber, with the scattered helium entering the detector chamber where it stagnates to form a stable pressure. By rastering the sample back and forth under the beam, an image of the surface may be constructed.

2.1.4. Gas Flow Model

The model of the flow of helium through the various chambers can now be explained, but again for simplicity the only system dealt with will be final prototype design wherein the major changes to geometry have been implemented. As a good example, the model originally had two differential stages prior to the sample chamber, as might be expected for a design originating with HAS systems. The extra differential stages were modelled in the same manner described below, but it was immediately found that there was a drastic loss in count rate due to the extra differential stage, and so it was removed. Also note that the base assumptions in the model attempt to replicate the worst case scenario with regards to the count rate to ensure viability of the instrument before construction began.

35 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

The behaviour of a supersonic free-jet beam source has been the subject of much work [34, 94, 95], leading to semi-empirically determined expressions for several useful parameters. The model makes use of these expressions to provide a simple estimate of the output of the source. In counts per second per steradian, the

centreline intensity (I) of the beam can be defined by the stagnation pressure (P0), nozzle diameter (d), the ratio of specific heats for helium (γ, 5/3 for helium) and

gas temperature (T0) using the following:

2 푃 푑2 퐼 = ( ) (퐶′ 0 ) , (Eq. 2.1) 휋 √푇0 where 1 1 휋 2 훾 2 2 훾−1 퐶′ = √ ( ) ( ) . (Eq. 2.2) 4 푘 푚 훾+1 훾+1

By then dividing this intensity by an empirical attenuation factor of 2.5 [96] we get a realistic measure of the nozzle throughput. For the prototype instrument using a 10 micron nozzle aperture, a room temperature beam and a stagnation pressure of ~200 bar, the model yielded a centerline intensity of 1019 - 1020 atoms/sec/steradian. The beam produced by the nozzle is then incident upon the sharp skimmer, which for the prototype would have a diameter of 100 microns. From the nozzle-to-skimmer separation and the skimmer diameter we can calculate the angle subtended by the skimmer, and hence the contribution of the supersonic free-jet into the differential stage. Note that the model was designed

with the nozzle-to-skimmer separation (ZNS) to be set at an ‘ideal’ value – one which would maximise the centreline intensity of the beam downstream. Campargue [94] gives an expression for this position, confirmed by Beijerinck [95], namely:

푑 푃0 1/3 푍푁푆 = 푐 푑 ( ) , (Eq. 2.3) 휆0 푃푏

where λ0 is the mean free path for the stagnation volume, Pb the background pressure in the expansion chamber, and c is a constant somewhere between 0.125 and 0.15, thought to be related to the skimmer form-factor. For the beam apparatus, with a 200 bar stagnation pressure this distance is between 9 and 12 mm (depending on the choice of c, temperature, mean free path, etc.). The axis between the nozzle and skimmer will be referred to as the Z axis for the source, with the other two axes being defined as X (horizontal) and Y (vertical). The model also assumes perfect alignment of the nozzle in X and Y with respect to the skimmer, a potential mechanism for a loss in helium signal should this not be the

36 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

case. Such an assumption was deemed acceptable due to prior experience with the alignment of nozzle assemblies with macro skimmers.

Any helium not passing into the differential stage is assumed to remain in the source chamber and immediately becomes part of the uniform background. By calculating the background helium pressure of the source chamber from this excess helium and the known pump rate for the chamber, the throughput of diffuse helium through the skimmer orifice can be estimated. This helium leak contributes to the background levels in the differential chamber.

The process is repeated for the differential stage. The size and position of the pinhole aperture sets the portion of the beam helium to pass into the sample chamber. As before, all other beam helium becomes part of the background in the differential stage, along with the leakage from the source. The conductance of the pinhole, the size of this background and the pump speed of the chamber then sets the contribution from differential stage to sample chamber. The size of this effusive contribution proved of particular importance to the microscope, and will be discussed in more depth later.

The remaining portion of the free-jet beam travels through to strike the sample surface. A true calculation of the scattering would be quite complicated and obviously totally sample dependent, so instead the worst case scenario is considered. Unlike a specular, or even a diffuse, reflection from the sample surface where the distribution is peaked, the model instead spreads the entire helium flux striking the sample surface into a uniform hemispherical cap (see Figure 2.5 below).

Figure 2.5 Representation of how the gas flow model handles the interaction of the helium beam with the sample surface. The same number of helium atoms incident on the sample from the free-jet beam are then spread evenly into a hemispherical cap centred where the beam strikes the sample.

Due to the interaction mechanisms of neutral helium with a surface [31], we can disregard the potential for helium sticking to the surface and not becoming part of the reflected population. As before, the amount of helium from this hemispherical

37 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

distribution through to the detector can be found knowing the distance and size of the detector aperture. Any excess helium not considered to move on to the next chamber once again becomes part of the background, and the model then determines the leakage from sample to detector.

The final step is to obtain the electronic signal registered by the quadrupole. The partial pressure is found for both the helium reflected from the sample surface (the signal of interest) as well as the diffuse background helium which leaks in from the sample chamber. It should be noted that the model calculates equilibrium partial pressures, meaning it assumes that enough time has passed to allow the helium population to stabilise. The pump rate in the detector is a controllable model variable, with smaller pump rates increasing the partial pressures, but also the time taken to reach an equilibrium. Based on experiments with another stagnation detector [97], pump rates were generally kept to somewhere in the 0 – 10 L/s range. With the signal and background partial pressures, the sensitivity of the Hiden quadrupole can then be used to convert them to a current and hence a count rate.

The model estimates the noise for the measurement as the square root of the total count rate (i.e.: shot noise [98, 99]). The noise floor of the Hiden quadrupole is also an issue to be considered, but at the time the model was being built it was in operation on another system and had been optimised thoroughly for helium. Due to this refinement of the mass discrimination, as well as the natural scarcity of helium (part of its commonplace usage as a leak test gas), the Hiden background count rate for a clean chamber was typically of the order of 5 Hz, and hence was safely ignored.

For a given instrument geometry, the Excel gas flow model could thus provide an estimate of the count rate, the signal-to-noise and the signal-to-background. In terms of the image quality goals discussed previously, it was found that for a design employing a 200 bar beam with a single differential stage and a 5 micron pinhole, count rates in the range of 2000 – 3000 per second were achievable. As such, provided that samples had at least 10% contrast available the desired image could be produced. Based on these results, preparations for construction were begun, culminating in the instrument detailed in Section 2.2 below. Although the model was initially built as a purely diagnostic tool, it was found that the predictions agreed quite well with the experimental results. Specific numbers from the model are discussed later as a direct comparison to the performance of the prototype SHeM.

38 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.2. Design and Experimental Setup

2.2.1. System Frame

As noted previously, the basis for the prototype SHeM was a pre-existing instrument built at the Cavendish laboratories for the purposes of testing technologies such as atomic mirrors [24, 60] and supersonic helium beam sources. Figure 2.6 below shows a render of that previous system and associated frame. The ‘Helium Source Chamber’ (as indicated in the diagram) became the source chamber for the new instrument, and the section of subframe it sat on was repurposed to mount the additional chambers. The other sections of the prior system were not utilised.

Figure 2.6 The previous vacuum system at the Cavendish laboratories, part of which was used to build the prototype SHeM. In particular, the source chamber and section of frame it rests on was repurposed for the new instrument.

39 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

The base frame consisted of welded steel I-beam sections, above which sat a subframe constructed from hollow square section steel. On top of the frame, tabletops (or in the case of the source chamber, a movable cart) could be sat to allow for the mounting of the chambers and pumps. The large mass provided by the frames is inherently resistant to vibrations, but the existing system had an additional mechanism to aid in this regard in the form of three airbags to damp vibrations predominantly in the vertical axis. However, based on the prototype nature of the instrument and predetermined pinhole size, it was deemed unnecessary to try to actively reduce vibrations, and as a result the airbags were left deflated to have to the frame rest directly on the ground.

2.2.2. Helium Source

With the chamber from the previous instrument as the basis, the helium source took the form shown schematically in Figure 2.7. The nozzle assembly was made in the style of Buckland et. al. [55], a design which has previously been employed in instruments both at Newcastle and Cambridge [97, 100]. As the majority of the design is commercially available Swagelok pieces, the construction was quick and low cost, in line with the design goals of the prototype microscope. Based in part on the pumping available to the source, provided by a Shimadzu TMU2203 2000 L/s turbomolecular pump backed by an Edwards E2M80 rotary vane pump, a nominally 10 um nozzle aperture was chosen for the source. The nozzle assembly was mounted to a simple L-bracket construction for support and to ensure alignment of the beam axis. The bracket was then attached to an X-Y-Z UHV- Designs manipulator allowing for precision control of the nozzle position relative to the skimmer. The source chamber reached an ultimate pressure of ~ 2 x 10-8 mbar, which typically rose to ~ 1.4 x 10-3 mbar during beam operation close to the maximum stagnation pressure of 200 bar (as dictated by the Swagelok and stainless steel pipework). During beam operation at 200 bar, the source was capable of producing (1.4 ± 0.1) x 1020 atoms/sec/steradian, as calculated from the source chambers exhaust gas flow rate.

40 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

41 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.8 Photos of the nozzle assembly during construction. Note the designated axes for movement of the nozzle.

The beam end of the source chamber was capped with a custom DN160CF flange on which the skimmer was mounted. The inner face of the flange was thinned out towards the centre to help minimise backscattered helium atoms (which cause beam attenuation and hence a loss in intensity) and to move the skimmer closer to the sample. The centre of the flange is bored through with a diameter such as to allow the skimmer (a Type 1 from Beam Dynamics with a 100 um central opening) to be easily clamped in place. The base of the clamping mechanism could be exchanged to move the skimmer position back and forth, allowing the travel range of the nozzle mounted to the manipulator to be altered if necessary. A cross- sectional schematic of the skimmer mount is shown in Figure 2.9.

Figure 2.9 Sectioned exploded schematic of the DN160CF flange on which the skimmer is mounted. The base of the clamp can be swapped out to change the travel range for the source nozzle relative to the skimmer.

42 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

The source chamber and its turbo pump were mounted on a cart so as to allow this section of the instrument to be pulled back, facilitating skimmer exchanges and alterations to the nozzle assembly. The skimmer itself is very delicate, and so this arrangement allowed it to be carefully and securely mounted externally, then placed into position with little risk of damage.

Figure 2.10 Photograph of the mounted skimmer prior to installation into the SHeM.

2.2.3. Differential Stage

The differential stage design proved to be a challenging part of the new instrument. With the premium placed on reducing the span of the beamline in order to bolster the helium signal at the detector, cutting the length of the differential stage was vital. Removing it entirely was not an option, as without the pumping, the helium background into the sample would be far too high to support imaging. With the space requirements dictated by the frame, and to keep fabrication simple, the differential stage became a section of DN40 tube welded into the DN160CF flange that formed one end of the sample chamber. Figure 2.11 is a schematic showing the orientation of the source, differential and sample chambers. The differential stage is pumped by a 520 L/s DN100CF Pfeiffer TMU521 turbomolecular pump backed by the same Edwards rotary pump as the source; however the turbo pump is connected to the differential stage via a short length of DN40CF bellows which restricts the available pumping.

43 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.11 Top cross-section of the prototype SHeM source (green), differential and sample chambers (blue). The differential stage consists of a DN40 tee welded to the front of the sample chamber flange, allowing for a turbo pump to be connected via bellows. The beam passes through the differential stage and onto the sample chamber by passing through the pinhole plate (purple), wherein the pinhole apertures out all but the desired spot.

2.2.4. The Sample Chamber

The sample chamber consists most importantly of the modified flange which forms the differential stage and detector connection. As it also holds the pinhole optics and the sample mount, it quickly becomes the most complicated part of the entire design. The rest of the chamber is cylindrical in nature (see Figure 2.11) with a Pfeiffer TMU521 turbomolecular pump mounted horizontally on the end, backed by a Pfeiffer DUO10M rotary vane pump. An ultimate pressure of 6 x 10-8 mbar for the chamber was achieved, although most images were taken with a pressure in the 10-7 mbar range as sample exchanges required bringing the chamber up to atmosphere.

44 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.2.5. Pinhole Optics

The pinhole is among the most critically important elements of the microscope design. The size of the pinhole determines the size of the beam spot that is projected onto the sample and thus controls the resolution of the instrument. Based on the modelling described previously, it was found that a pinhole of approximately 5 micron would be a good starting point in terms of balancing the count rate and resolution, although sizes down to 1 micron should be feasible (albeit with an increased imaging time). Apertures of this size range as used in optics are readily available, however the 45° geometry poses a complication: if the pinhole material is thicker than the size of the pinhole, then line of sight from beam source to sample may be lost. This idea is demonstrated in Figure 2.12 (a). One way around this issue is to mill the pinhole at an angle through the substrate, as in Figure 2.12 (b). Such an approach was trialled using laser milling of 50 micron thick stainless steel discs, but the process resulted in misshapen pinholes, and alignment of the disc to the actual beam axis was deemed too difficult to be a practical solution. Instead, it was decided that field ion beam (FIB) milling through a resilient membrane would provide the required pinhole for the instrument. Figure 2.12 (c) shows this solution, where the membrane (surrounded by a much thicker support structure) is thinner than the pinhole, allowing the beam to pass through unobstructed.

Figure 2.12 Illustration of the potential ways to implement a pinhole at 45 degrees to the beam axis (represented by the green arrow). If the pinhole is added normal to the plane of the membrane as in (a), then unless the membrane is thinner than the width of the pinhole line of sight to the sample is lost. In (b) the pinhole is bored along the beam axis, but manufacturing and alignment concerns make this difficult. (c) displays the compromise utilised in the prototype SHeM: A pinhole is bored through a very thin section of material, allowing the beam to pass through as desired. In order to implement this practically, the pinhole substrate was chosen to be a silicon nitride disc with a small membrane section in the center, thus allowing it to handled and glued into position.

45 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

To maintain a pinhole that could still be handled and mounted successfully, silicon nitride discs from Ted Pella (part number: 21525) with a 250 x 250 micron, 200 nanometre thick, square membrane were employed. A thin layer of platinum was applied to the membrane prior to milling to prevent charging effects.

The silicon nitride discs were glued using vacuum leak sealant into a shallow depression in an aluminium plate (henceforth referred to as a ‘pinhole plate’, see Figure 2.13) which had conical pathways for both the incoming beam and the reflected helium. A second hole in the pinhole plate provided the entry for the latter. The pinhole plate could then be bolted to the inner wall of the sample chamber (as in Figure 2.11) allowing for easy exchange. Viton O-rings slotted into the back of the plate ensured good seals to the inner wall, helping to better isolate the differential stage from the sample chamber, and the sample chamber from the detector volume. It should also be noted that the design of the pinhole plate sets the distance from pinhole to sample along the beam axis (the working distance) for the instrument. For all the experiments detailed in this chapter, a pinhole plate with a working distance of 3 mm was utilised.

Figure 2.13 Sectioned render of the pinhole plate used to hold the silicon nitride disc with the pinhole bored through the center. The channel to the right shows the beam entry pathway, including the shallow depression on the face of the plate where the silicon nitride disc is located. The second pathway is to allow reflected helium to travel on to the detector chamber. The point where the axes of both channels meet (as shown in the inset) is the sample specular position.

46 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.14 Photograph and optical micrographs of the 5 um pinhole implemented in the prototype SHeM. The silicon nitride disc was glued in place in the pinhole plate with vacuum leak sealant to provide a strong bond and stop extraneous gas escaping around the edges of the disc. As can be seen in the micrographs, the 250 micron square film has the pinhole bored through its center. The rippling of the membrane was caused by the field ion beam milling process.

2.2.6. Sample Mount and Piezo Drives

A stainless steel sample slide, trapezoidal in shape (see Figure 2.15), was designed to slot between three phosphor bronze pins mounted in a base plate. When the pin assembly is sat vertically, the sample slide can be lowered down to firmly lock into position with minimal force. The mechanism was found to provide excellent repeatability in terms of the sample position. To alter the distance of the sample from the specular position (henceforth referred to as the Z-position of the sample), a simple jackscrew could be adjusted to move it relative to the pinhole plate. The pin assembly itself was attached to two Attocube ECS3030 piezo slipstick drives arranged on top of each other, allowing it be rastered underneath the beam. With a 1-5 micron pinhole, the sample would be required to be stepped in at least micron increments, and the potential for larger samples (perhaps up to a square centimetre) necessitated a large travel range. Furthermore, the confined space within the sample chamber close to the pinhole plate imposed limits on the possible footprint of the drives. The Attocube ECS3030 drives satisfied the footprint and travel requirements (the 30 x 30 x 9.5 mm units are capable of nanometre positioning over a 20 mm travel range), were able to move a load of 90N, and were equipped to be UHV compatible. Due to manufacturer availability (the UHV

47 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

variants being prototypes themselves), the drives were not equipped with absolute position feedback and thus ran as an open loop system.

Figure 2.15 Clockwise from top left: Sample slide with Z-distance screw in place and a TEM grid mounted, ready for imaging; Sample mount mechanism assembled out of the chamber (note the designated labels for the movement of the sample); one of the Attocube ECS3030 piezo slipstick drives responsible for the rastering of the sample underneath the beam; the sample mount mechanism assembled and bolted to the pinhole plate, showing the typical imaging position.

Control of the drives was achieved via a PC with a Matlab script communicating with the control unit. Several different scan routines were written, allowing specification of the scan area, pixel count, step size, dwell time per pixel, (etc.) for a number of scan patterns. The piezo stages, attached to a base plate, were affixed to the edges of the pinhole plate with three legs and a simple spring mechanism to provide stability. With no attempt at actively minimising vibrations through the instrument, mounting off the pinhole plate ensured that if there was any significant movement the sample would move in sync with the pinhole.

48 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.16 Sectional top view of the sample mount (red) along with the pinhole plate (purple) as mounted to the inner wall of the sample chamber (blue). The incident beam axis is marked in dotted lines, with the specular position (and hence the sample location) at the point where it crosses the axis to the detector.

2.2.7. The Detector Chamber

With the necessary restriction of operating a detector in stagnation mode, the principle concern for the detector chamber became to house the Hiden Analytical HAL/3F-PIC quadrupole head with as little extra volume as possible. As discussed in Chapter 1, operating the detector in stagnation mode with reliable results is dependent on the pressure in the chamber rapidly reaching a stable equilibrium. As the rise time is dependent on the chamber volume, minimising this volume reduces scan times by cutting down on the dwell time prior to sampling and improves the reliability of the measurements. The smaller volume also means that (in general) a better ultimate pressure will be achievable, boosting the signal to background ratio for the instrument.

The net result of the requirements was a chamber which acted as a tight sleeve for the Hiden quadrupole head. A pumping port was added as close to the front of the quadrupole as standard DN16CF fittings would allow to ensure it reached an equilibrium population. To further ensure the detector would be clean, especially considering the adjacent sample chamber would be brought up to atmosphere during sample exchanges, a gate valve was added between the sample and detector chambers. This addition allowed the detector chamber to be sealed off and allowed to be kept pumping at all time, with repeated bakes also utilised to

49 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

keep the background low. Below shows a cross-section of the instrument highlighting the pathway of the helium from sample to detector.

Figure 2.17 Cross-sectional side view of the sample and detector chambers highlighting the path of the reflected helium atoms from the sample through to the Hiden quadrupole. A 1 mm diameter aperture in the pinhole plate admits a portion of the specularly reflected helium atoms, which then pass through the custom sample chamber flange, onto the detector chamber proper. A stable population builds, which is sampled by the quadrupole.

The pumping available to a detector operating in stagnation mode is critical – too much pumping and a large enough population will not build, and too little and the pressure will not reach equilibrium in a reasonable amount of time. When setting up the microscope, a long length of DN16CF bellows was used as a temporary connection between the pumping port on the detector chamber and its Pfeiffer TMU521 turbomolecular pump, with the aim being to replace this with something with a larger conductance and a copper gasket with a hole to control the available pumping. However, it was found that the length of brought the conductance down to an estimated 2 L/s, yielding a good balance in the pump speed, and so it was retained.

The Hiden quadrupole was controlled via a PC with the Hiden software suite MasSoft allowing the various parameters (mass selection, filament current, etc.) to be set. Pulling the count rate per pixel during the scan routines was accomplished by connecting the output of the Hiden RF head to an Agilent pulse counter (model 53131A), the output of which was fed into a separate PC. The same Matlab scripts moving the piezo stages in a raster pattern would pull the count rate for a designated dwell time and write the values to file. With the detector operating in stagnation mode, the scan software critically allowed for both a wait time and a dwell time to be specified.

50 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.2.8. Beam Path Length

As discussed, the dimensions of the microscope which have the most impact are those which, taken together, give the path length of the helium beam from nozzle to detector. For future reference, the following are the distances which make up the beam path of the prototype SHeM:

 Nozzle-to-Skimmer Distance: Based on the previously mentioned calculations of the ideal nozzle to skimmer separation, the nozzle was most commonly positioned at 10 mm (the lower end of the calculated range, where count rates were found to be higher). As was found to be case during experiments, pushing the nozzle closer did increase the count rates at the detector. However, the helium partial pressure in the differential stage rose beyond what the pumping could adequately handle, leading to imaging problems. This issue is discussed in depth in Section 3.2.

 Skimmer-to-Pinhole distance: Retaining a differential pumping stage in the design was required to keep the background to a manageable level, despite the concerns over the additional length it would incur. The distance between the skimmer and pinhole apertures was chiefly set by the need to incorporate a pumping port on the tube connection that makes up the differential stage. In the end, this stretch of the beam path was 140 mm in length.

 Pinhole-to-Sample Distance: The working distance was set by the design of the pinhole plate. In particular, moving the pinhole and detector apertures closer together would drop the working distance (and increase the count rate significantly), but start to cause problems with the silicon nitride disc starting to run afoul of the detector aperture. The 3 mm separation from pinhole to sample along the beam axis was a compromise between the count rates, and a part that the Cavendish workshop could build in the required time frame. The 5 micron pinhole utilised in the prototype instrument subtended an angle of approximately 2.2 x 10-6 sr at the nominal imaging position.

 Sample-to-Detector Distance: With the 45 degree geometry, the aperture through from the sample chamber to the detector is placed at the same distance from the sample as is the pinhole aperture, namely 3 mm. Once helium has passed this point, the nature of stagnation detection means that length is not as much of an issue as in the previous sections. That said, efforts were taken to minimise the internal volume within which the detector sits, and hence the length from the back of the pinhole plate to the quadrupole head was kept as short as fittings would allow. The 1mm detector aperture in the pinhole plate subtended an angle of approximately 8.5 x 10-2 sr at the nominal imaging position.

51 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.2.9. Completed Microscope

The figures in the following section show the completed microscope in operation at the Cavendish laboratories. Also included (Figure 2.21) is a gas flow schematic of the SHeM, as well as the beam control panel. The panel, along with its booster and cold trap, were taken from the previous helium beam that operated in the salvaged source chamber.

Figure 2.18 Photograph from the side of the prototype instrument in operation, with the source chamber on the left and the sample chamber with its turbo pump to the right. Also visible is the cart the source chamber sits on, allowing it to be moved back from the rest of the chambers to provide access for skimmer interchanges and adjustments to the nozzle assembly.

52 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.19 Photos of the completed instrument during operation. The top photo shows a top down view of the beam path from source to sample to detector (the latter wrapped in foil for baking purposes at the time). The bottom photo looks back past the detector chamber towards the sample and source chambers.

53 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.20 Photograph of the front of the beam control panel, showing the regulators, valves and gauges used when supplying the compressed helium to the nozzle. Behind the panel face sits the Swagelok pipework (see schematic below). Additionally. you can see the helium cylinder to the left, the gas booster at the bottom center, and the dewar which comprises the outer shell of the cold trap to the bottom right.

54 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

Figure 2.21 Gas flow schematics of the beam control panel and prototype SHeM. High vacuum stages are shown in black, rough vacuum in blue, high pressure components in red, and gas storage in green. The gas panel allows bottled helium to be compressed to the neighbourhood of 200 bar and then regulated down to the desired stagnation pressure. It should also be noted that the helium was passed through a cold trap in order to filtrate any remaining impurities. Bypasses link the chambers and thus prevent the possibility of damage to the skimmer or pinhole when bringing the system up to atmosphere (or the reverse)

55 CHAPTER 2: The Mark I Prototype SHeM – Design Overview

2.3. Conclusions

Based on encouraging results from initial attempts at modelling key aspects of the system, a prototype scanning helium microscope was assembled at the Cavendish laboratories in Cambridge. In order to achieve a working instrument, the primary goal of the design was to significantly reduce the path length from nozzle to detector in order to maximise the available helium signal. While posing interesting design challenges with regards to differential pumping and sample mounting, the reduced beam path length did facilitate the use of a simple pinhole as the final optical element. The following chapter goes over the experimental studies carried out on the new instrument, including characterisations of the source and downstream optics, as well as dedicated imaging studies.

56 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

CHAPTER 3

THE MARK I PROTOYPE SHEM – EXPERIMENTAL RESULTS

In this chapter we describe the experimental results obtained with the prototype SHeM detailed previously. The performance of the beam source is assessed, followed by a discussion of one of the key discoveries associated with the instrument geometry – the presence of a secondary effusive helium beam. Found to be detrimental to the quality of the produced micrographs, the strength of the secondary beam had to be lowered before instrument performance replicated that predicted by the modelling. Typical SHeM micrographs are presented for the optimised imaging parameters, the perspective offered by the scattering geometry examined, and some select materials systems are investigated for reasons of contrast. Finally, an evaluation of the design performance is made, with a list of alterations to be implemented in further iterations of the instrument. Data collection for the Mark I SHeM was completed in collaboration with Matthew Barr, Chris Wade and Dr. Paul Dastoor; all subsequent data analysis was performed by the author.

3.1. Source Characterisation

The first step to ensure the instrument was performing as expected was to evaluate the helium beam source. Once the Buckland style nozzle assembly [55] was confirmed to correctly seal, the aperture performance had to be checked to ensure that, for example, it had not been clogged by any stray particulates during

construction. The source chamber pressure Pb should be linearly related to the

stagnation pressure P0 for a correctly functioning supersonic free-jet expansion beam source as given by the equation

2 20 푘 푇 ′ 푑 푃푏 = [( ) (퐶 )] ∙ 푃0 , (Eq. 3.1) 휅 푆 √푇0

where k is the Boltzmann constant, T the chamber temperature, T0 the stagnation temperature, d the nozzle diameter, S the chamber pump rate, C’ as defined in

57 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

equation 2.2, and κ the so called ‘peaking factor’ (for helium, approximately 2 [101]). Figure 3.1 shows the collected experimental data along with a linear fit, the quality of which (R2 value of 0.997) indicates the source was behaving correctly. The slope of the linear fit was then used to determine the nozzle diameter, yielding a value of (10.08 ± 0.24) microns, in good agreement with the nominal diameter of the aperture (10 um).

Figure 3.1 Plot of the corrected source chamber pressure as a function of beam stagnation pressure for the prototype SHeM. The linearity of this data, as shown by the quality of the linear fit (R2 value of 0.997), demonstrates the expected performance for a supersonic free-jet beam source. From the slope of the fit – (3.81 ± 0.18) x 10-5 – we derive an estimate for the effective nozzle diameter of (10.08 ± 0.24) microns, in good agreement with the nominal diameter of 10 microns.

The chamber pressures predicted by the linear fit in Figure 3.1 match those found experimentally up to the maximum stagnation pressure used in the source (200 bar). The helium partial pressure in the source chamber for a 200 bar beam was (8.00 ± 0.50) x 10-3 mBar, and using this value we can calculate the centreline intensity of the beam via the exhaust gas flow rate by means of the equation:

휅 푆 퐼 = ( ) ∙ 푃 . (Eq. 3.2) 0 10 휋 푘 푇 푏 For the Shimadzu TMU2203 capable of pumping 1200 L/s for helium, we calculate the centreline intensity of the room temperature helium beam to be (1.48 ± 0.10) x 1020 atoms / second / steradian. Such a value is in excellent agreement with the predicted beam intensity from the prior gas flow modelling (see Section 2.1.4).

58 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Next in the installation was to verify that placing a surface in line with the beam in the sample chamber would reflect helium towards the detector aperture. Firstly, the helium signal as detected by the Hiden quadrupole was recorded for a series of nozzle-to-skimmer separations with no sample in place. The experiment was then repeated with a blank sample slide in the sample chamber, meaning that the beam was incident on the top of the sample stud (see Figure 2.15). The collected data is shown in Figure 3.2.

Figure 3.2 Helium signal as detected by the Hiden quadrupole as a function of nozzle-to-skimmer separation both with and without the sample stud sitting at the specular position. Scans were performed with a stagnation pressure of 120 bar at room temperature, and a pinhole plate without pinhole installed between the differential and sample chambers. The difference in signal is due to the helium beam reflecting off the metal surface into the detector aperture.

It should be noted that the experiments were performed early on in development with a pinhole plate without a pinhole glued in, meaning the final aperture for the beam was a 2 mm diameter hole. The count rate as observed by the detector was seen to clearly increase with the addition of the sample slide, indicating helium atoms interacting with the metal surface at the specular position were making their way into the detector volume. Similar experiments were performed once a silicon nitride membrane with a 5 micron pinhole had been added to the pinhole plate, including driving the sample in and out of the beam path to confirm detector response. It should be noted for the purposes of future discussion that at the time

59 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

the pinhole was installed, the nozzle-to-skimmer separation was reduced as much as possible to maximise the count rate.

With the helium beam source performing to specification and the instrument geometry confirmed to produce a sufficient detected helium signal after the beam interacts with the sample surface, an attempt to collect an image by rastering the sample underneath the beam was made.

3.2. Effusive Beam Contribution

Initial attempts to produce images of samples with known features, namely the bars of a TEM grid, were unsuccessful. The micrographs produced showed little in the way of contrast and did not suggest the known surface topography. The results were indicative of a significantly larger beam spot on the sample surface (as compared to that expected for a supersonic free-jet expansion apertured with a 5 micron pinhole), a hypothesis revealed to be correct after further experimentation showed evidence of a secondary effusive beam. With the limited pumping on the differential stage of the instrument, the helium partial pressure increased to the point that the mean free path of the helium atoms became comparable to the size of the pinhole. As such, the chamber acts as a Knudsen cell, producing an effusive beam from the pinhole, which then competes with the supersonic beam originating from the nozzle.

The problems that the secondary beam causes stem from both its intensity, as well as its shape. The angular distribution of gas for an effusive beam depends on the Knudsen number K for the environment of the expansion, as well the nature of the channel as defined by β (see Chapter 1). If free molecular flow is maintained (i.e.: K >> 1) then only collisions with the wall of the channel occur (making it ‘transparent’) and the flow properties are determined solely by the geometry of the channel – that is, by β. For a thin walled orifice, the value of β → ∞ and the effusive beam will take on a cosine angular distribution. As the value of β gets smaller, the beam will become more directional as can be seen in Figure 3.3a. However, if the channel is ‘opaque’ (i.e.: intermolecular collisions are of comparable importance to wall collisions; typically quoted as values of K ≤ 10) then the shape is dependent on both K and β [51]. Figure 3.3b shows the change in angular dependence for a β of 0.05, showing the increase in directionality for larger K values.

60 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.3 Theoretical angular distributions for effusive beams under different conditions. (a) Angular dependence of the beam profile on the value of β for a source with a Knudsen number K >> 1. β = ∞ refers to the case of a thin walled orifice, which then causes the effusive beam to follow the familiar cosine distribution. (b) Demonstrates the change when K ≤ 10, and the shape becomes dependant on both K and β. For the given β value of 0.05, the directivity of the beam source can be seen to increase with larger Knudsen numbers. Figure from [51]

For the Mark I SHeM, the pressure downstream from the differential stage during beam operation rose to a maximum of ~ 5 x 10-6 mBar. With the shape of the pinhole plate and the position of the differential stage ion gauge, the partial helium pressure around the pinhole was almost certainly much higher. Taking the worst case scenario for the source of the effusive beam, we estimate the pressure in the vicinity to be an order of magnitude higher - making the mean free path around 4 metres. Even if the mean free path were smaller than this, the width of the silicon nitride membrane means that the Knudsen number at the pinhole is much greater than 1, and so the shape of the produced effusive beam is based purely on β. A 5 micron pinhole through the 200 nm membrane yields a β value of 25, and thus based on Figure 3.3a we see that the produced effusive beam will be very close to a cosine distribution in shape - a marked difference in shape to the primary free jet beam.

The presence of the secondary beam will act to degrade the quality of the images produced by the instrument in several ways. The effusive beam represents a leakage of helium from the differential stage into the sample chamber, and a higher background in the sample chamber will reduce the signal-to-background ratio. However, the effusive beam is not simply a leak – it is a directional leak. While the

61 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

cosine shape of the distribution will cause it to form a broad background term with little variation across the sample surface, it is striking a large portion of the sample surface simultaneously. As a consequence, areas far from the point of impact of the supersonic beam are contributing to the collected helium signal for that particular pixel and, as such, the available contrast will be seen to diminish. For example, if we were looking at a section of sample surface that should appear ‘black’ in a SHeM micrograph (indicating a large scattering away from the specular direction, or some form of occlusion), the effusive beam would strike neighbouring areas which would then ‘brighten’ the pixel by incorrectly increasing the particle count. Thus, the broad effusive beam can be seen to wash out the contrast in a collected image, bringing white and black areas of the image closer together in intensity. With the resolution of the instrument not just being dependant on the beam shape, but also on the level of contrast available with which to resolve features, the loss of contrast can also be considered a loss of resolution. Further discussion of the deleterious effects due to the effusive beam appears in Chapter 5 with regards to the second generation SHeM, where experiments were carried out specifically to demonstrate its effects.

In order to minimise the contribution of the effusive beam to the produced SHeM images, the helium pressure in the differential stage had to be reduced. As increasing the available pumping was not possible, the only alternative was to vary the position of the nozzle relative to the skimmer. If the nozzle is close to the skimmer, the angular acceptance of the latter aperture is maximised and thus a greater amount of helium is able to pass into the differential stage. Pulling the nozzle back from the skimmer will then reduce the differential stage pressure and correspondingly weaken the effusive beam. By scanning the nozzle position across the face of the skimmer in one axis for a series of nozzle-to-skimmer separations, the strength of the two different beams can be visualised (Figure 3.4). It should however be noted that the profiles taken in this manner are not explicitly profiles of the helium beam, but rather a more complex measurement of instrument geometry. The scans shown in Figure 3.4 were once again taken with a blank sample slide, and the beam stagnation pressure set to 100 bar.

62 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.4 Detected helium signal as a function of varying nozzle-to-skimmer separations. At small separations, the pressure in the differential chamber will increase such that the effusive beam will dominate the free-jet beam, producing a broad peak unsuitable for imaging. However, by pulling the nozzle back from the skimmer the differential stage pressure drops and the sharp free-jet beam emerges. Note that the nozzle-to-skimmer separations have a zero offset of up to 1 mm due to the difficulty of aligning the nozzle with the fragile skimmer.

Figure 3.5 Comparison of the differential stage pressures as nozzle position is varied at the extremes of the nozzle-to-skimmer separation from Figure 3.4. At 10mm separation, the helium partial pressure is an order of magnitude smaller than that at the 5 mm separation (1.4 x 10-5 mBar vs. 1.7 x 10-4 mBar in the central position).

63 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

As the plots in Figure 3.4 and Figure 3.5 demonstrate, if the nozzle-to-skimmer separation is too small then the differential stage pressure gives rise to an effusive beam which will dominate the primary beam. Even when the nozzle is drastically misaligned in relation to the skimmer (meaning the centre of the supersonic free- jet expansion is not passing through the pinhole), there is still a significant detector count rate from this effusive beam. Pulling the nozzle back brings the partial pressure of helium down significantly – at 5 mm separation, the differential stage ion gauge yielded a helium partial pressure of 1.7 x 10-4 mbar, versus 1.4 x 10-5 mBar at 10 mm separation. Consequently, the sharp free-jet beam emerges and the effusive beam is reduced to a broad background. Raising the stagnation pressure from the 100 bar used to obtain the data in Figure 3.4 and Figure 3.5 to the 200 bar used for typical imaging conditions increased the differential stage pressure further, but the free-jet beam also intensifies. It was found in practise that keeping the nozzle around 10 mm from the skimmer resulted in a good balance between signal and effusive beam size.

Interestingly, the requirement to increase the nozzle-to-skimmer separation to at least 10 mm also places it in the ‘ideal’ region defined by Campargue and Beijerinck [94, 95]. Evaluating equation 3.2 for the experimentally determined parameters (i.e.: a nozzle diameter of 10.08 microns, a stagnation pressure of 200 bar, a background pressure of 7.5 x 10-3 mBar and a corresponding mean free path of 5.85 x 10-10 m), the calculated separation is between 9.7 and 11.7 mm (depending on the value of c). It should be noted that the theoretical ‘ideal’ region centers around minimising the amount of back reflections of particles from the skimmer, which then collide with other atoms in the free-jet expansion and thus attenuate the beam. Just as with virtually every other application of atomic and molecular beams, when attempting to maximise the recorded intensity, nozzle position is crucial.

Once the change in nozzle-to-skimmer separation had been made and the differential pressures reduced, it was found that recognisable images could be generated. For example, Figure 3.6 shows some of the first SHeM micrographs the instrument produced of TEM grids, demonstrating the difference the simple change made to image quality. Once this point was reached, the prototype SHeM was able to move beyond initial characterisation to full imaging of samples.

64 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.6 The SHeM micrograph on the left shows a region of a TEM grid before the nozzle-to-skimmer distance was increased, while that to the right was taken afterwards. The appearance of the expected grid pattern with the increase in distance demonstrates the clear difference in instrument performance once the secondary effusive beam contribution is minimised.

3.3. Optimised SHeM Imaging

With any new microscopy, ensuring that the micrographs produced match the sample being imaged is of obvious importance. Therefore, the majority of the initial experimentation performed with the prototype was on a known sample surface; namely copper TEM grids. The grids provided a series of regular, well ordered features of known size with high contrast under the helium beam (due to the gaps through which the beam may pass). The grids are rough on a length scale well below the beam spot size of the instrument when employing a 5 um pinhole, and so one would expect fairly uniform diffuse scattering across the surface of the

65 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

sample. The grids were attached to the sample stud using carbon tape; a vacuum safe adhesive comprised of a carbon fibre matt impregnated with an acrylic glue. The grid appearing in the micrographs within this chapter had bars with a 24 micron width and a periodicity of 84 microns. Figure 3.7 shows one of the first grid images recorded (32 x 32 pixels in size), matched to optical microscopy of the same area.

The following discussion details the scan routine used to collect the intensity per pixel in the presented micrographs. In the Matlab software built to control the drives, the user could designate both a ‘wait’ (the length of time the instrument pauses at a new position) and a ‘read’ time (how many seconds the detector collects for) per pixel. Increasing the wait time allows the pressure in the stagnation volume to better approach the equilibrium value and provides a more accurate reading of the relative intensity, while increasing the read time collects a larger signal and hence improves collection statistics (e.g.: signal-to-noise). Henceforth, if the total time per pixel is discussed it will be referred to as the ‘dwell’ time. As an example, the SHeM micrograph in Figure 3.7 was recorded with a wait time of 0.5 seconds and a read time of 4 seconds, giving a dwell of 4.5 seconds. For the 32 x 32 pixel image, this yielded a total scan time of around 90 minutes, noting that there is some overhead in the software when travelling between pixels, recording data to a text document, and initial start-up.

As shown in Figure 3.7, the SHeM micrograph is extremely well matched to the optical micrograph of the copper grid, demonstrating that the produced image is an accurate representation of the sample geometry. Based on the scan routine and method of data collection, the beam strikes the sample from the right of the image, with the detector to the left. As the beam axis relative to the sample influences the produced micrograph, this convention will be kept for the rest of the chapter, unless specified otherwise.

Several features of note are apparent even in this early image. Square patches to the left of the micrograph in Figure 3.7, corresponding to the visible areas of the stainless steel sample stud, are quite dark (indicating that little helium from this point on the sample surface was reflected into the detector). The stud is the base of the sample, and so is the furthest point from the detector in the micrograph. A plane further away from the detector (that is, lower down in the assembled sample slide) will see a smaller angle subtended by the detector aperture. As the angular acceptance is smaller, if the scattering from two surfaces is equal then the lower surface will appear darker in the produced micrograph. The lack of signal in the region at the left of the micrograph could also be attributed to the grid on top

66 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

occluding some of the reflected helium atoms from the stud from reaching the detector. Such a process can be directly observed in another feature in the SHeM micrograph; namely the darker areas which appear in the bottom left corners of the grid situated over the carbon tape (black material in the optical micrograph under the majority of the grid). At these positions, the incident beam strikes the carbon tape and reflects, but as it travels towards the detector it intercepts the side of a spar of the TEM grid. Thus, at positions close to the raised grid surface (relative to the carbon tape), the intensity is reduced, producing masking effects. Due to the slight rotation of the grid axes relative to the beam-detector plane, the effect is observed on the left and bottom edges of each gap in the grid, with local variations in the amount of masking depending on the relative height of the carbon tape to the top of the grid (i.e.: if the relative height difference is bigger, the size of the masked area will be larger). Such occlusion effects are intrinsic to the scattering geometry of the instrument and thus critical to understanding the produced images, and will be discussed in greater detail in Section 3.4.

Figure 3.8 SHeM micrographs of a TEM grid demonstrating the problems with drift in the horizontal axis due to a problem with the translation stage. In some instances the stage would drift initially but then work as expected (Figure 3.7 is such an example), while in others it might continually drift throughout the entire scan (left image). The most detrimental type of drift was an inconsistent variation throughout the scan (right image), making post-correction difficult. Note in the right hand image, the transition between the bright and dark areas underneath the TEM grid correspond to the carbon tape / sample stud interface as described for Figure 3.7.

The next series of images taken with the SHeM (a sample of which can be seen in Figure 3.8) revealed a fault of the Attocube stages controlling the sample position resulting in significant amounts of drift in the horizontal axis. While the horizontal bars of the grid were quite straight, the vertical bars were warped, meaning the horizontal Attocube stage was losing its position. The problem was initially thought

67 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

to be temperature related with significant drift found at the start of scans (as the stage temperature rose during operation), and either much less or no drift occurring if multiple scans were performed in quick succession. While extensive adjustments to the control of the questionable stage were trialled in an attempt to resolve the issue, the problem was found to be a physical defect in the friction layer in the slipstick drive. Consequently, the behaviour persisted until a second unit was provided by the manufacturer. As a result many SHeM micrographs show warping, which in addition to the visual distortion also makes applying scale bars difficult (or impossible).

Figure 3.9 Schematic depicting horizontal drift causing image artefacts. Instead of the expected scan path shown in (a), the problems with the Attocube stage controlling horizontal motion would cause it to drift further in one direction, resulting in the scan path to instead be that shown in (b).

Figure 3.10 Two SHeM micrographs of adjoining regions of a TEM grid taken in succession. The lack of closed loop control in the scan stages leads to the mismatch in horizontal starting position.

68 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

The other consequence of the drift in the horizontal sample stage, along with the lack of closed loop control, was the inability to reliably find specific features for further imaging once points of interest appeared in initial scans. While the general area could be found, precisely imaging a desired region generally required much trial and error in setting up the scan. In practise, precision was limited (especially in the horizontal axis) to ~20-40 microns depending on the current state of the drives. By way of example, Figure 3.10 shows two micrographs taken in succession of adjoining areas of a TEM grid. The scans should align vertically, but as can be seen by the degree of offset required to match the spars, the starting point is some 10 microns from that desired.

3.4. Shadowing and Masking

The interesting relationship between the incident beam, sample and detector position that the 45 degree geometry affords presents some challenges when beginning to interpret the produced micrographs. As such, well-understood samples are a priority in order to relate the results to the nature of image generation for the emergent microscopy. To begin an investigation of the process, let us consider the instances where there will be occlusion of either the beam or the detector. Taking some conventions from micro-facet theory (used in the modelling of, among other things, diffuse scattering of light from surfaces [102, 103]), we can divide occlusion into two categories: shadowing and masking. Shadowing occurs when a section of a surface is not visible to the incident beam direction (and thus not contributing to the reflection response), while masking occurs when the section is not visible in the view direction (in our case, not within line-of-sight of the detector). Note that these definitions and the immediate discussion to follow neglect multiple scattering events due to their inherent complexity.

The SHeM micrographs in Figure 3.7 and Figure 3.8 have masked areas where the helium beam has reflected off the lower planes of the sample, but then struck the side of a TEM grid spar in transit to the detector. With the beam entering from the right of the micrograph, the masked areas will also appear to the right of the feature which causes the occlusion – for instance, the dark patches on the right- hand side of each TEM grid spar. To then understand where the shadowed regions appear in the same micrographs, consider the schematic of the beam interaction with a simple sample as shown in Figure 3.11, namely a flat plane with a large asperity. As the sample is scanned across during imaging, the protrusion begins to interfere with the beam and cause masking as in (b). Once it moves far enough

69 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

across, the beam will now strike the asperity before the sample, leading to shadowing (c). However, despite the detector ‘looking’ at a point well past the feature, any of the helium reflected non-specularly from the asperity will form the signal for that pixel. As the point on the sample the beam was intended to strike is not visible to the incident beam, such an arrangement still constitutes shadowing, but will appear as the image of the asperity in the produced micrograph. As such, in the TEM grid micrographs so far shown, the spars of the grid will not produce shadows – rather, they are the shadows. Figure 3.12 illustrates beam positions for both masking and shadowing with the aid of a simple three-dimensional form as the sample under investigation.

Figure 3.11 Simplified 2D illustration of masking and shadowing in the SHeM. Image (a) shows a simple sample with a large asperity being scanned under a helium beam (green) as shown by the arrow. When the beam strikes the sample sufficiently far away from the asperity, it makes no impact on the collected helium signal. However, in (b) the sample has been scanned across, leading to the reflected signal from the surface striking the protrusion, meaning that an area is masked due to detector occlusion. Further on at (c), the asperity is now blocking the beam from illuminating a section of the sample surface (shadowing). Note that part of this shadowed region will include the top of the asperity – provided the detector acceptance angle is sufficient, helium striking the top will enter the detector, forming the image at this position. Finally, the asperity moves far enough to no longer impact the incident beam as shown in (d).

70 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.12 Trimetric projection of a simple geometric sample being imaged via SHeM. Red arrow indicates the beam position for a masked area, which results in the cross hatched region of lower intensity. The blue arrow indicates a position where the substrate is shadowed by the sample - in fact, each pixel showing the sample will constitute a shadowed region where the beam was prevented from striking the substrate. Image to the right displays the resultant micrograph.

In practise, any structure on a sample which might be considered an asperity is part of the sample in question, and is likely much more complicated in geometry than that shown in Figure 3.11. As a result, drawing a distinction between what is masked and what is shadowed is generally only useful when explaining image formation within the SHeM. It should also be noted that although areas which are masked or shadowed will usually record lower helium intensities, effects such as multiple scattering, the width of the beam and detector aperture, and background helium gas will all contribute some intensity for those pixels.

To further expand on how the geometry affects the produced images, we can consider projections of the sample surface along both the beam and detector axes. If we project the area of the sample visible to the beam down onto the sample plane, the result is the full image as collected by the detector, with the exception of the masked regions. One can repeat the process from the detector perspective – everything not visible will be masked in the final image. Figure 3.13 shows both projections for a simple needle sitting on a flat plane, for a variety of angles between the needle and the sample plane (θ). As θ varies, the relative amount of shadowing and masking present can be seen to vary significantly.

71 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.13 Illustration of the consequences of the incident beam geometry on the produced SHeM micrographs. The helium beam enters from the right at an angle α relative to the normal for the base plane (for the Mk I SHeM, α=45º), and interacts with a simple sample (a needle). Specular reflections from the sample and base plane are shown via the red arrows, with the detector positioned to the left. Yellow indicates a shadowed region, while blue represents masking. As θ (the angle between the needle and the base plane) increases from 0º through to 180º as shown in the progression from (a) through to (e), the apparent length of the feature (shadowed region) can be seen to change significantly, a phenomena known as projection distortion. When the needle sits normal to the base plane as in (c), the 45º incident beam results in the apparent length equalling the true length. Note that once θ moves past 135º (as in (e)), the shadowed region appears to reverse direction, with the far side of the needle now exposed to the helium beam.

72 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

The series of images in Figure 3.13 demonstrates a phenomena known as ‘projection distortion’, whereby the apparent length of a feature in a produced micrograph can vary as a function of the relative angle with respect to the beam (or equivalently, with respect to the sample plane). As shown in the figure, unless the feature is parallel to the base plane the beam incident at 45 degrees can cause its apparent length (the size of the shadowed region) to appear longer or shorter than the true length. This is in contrast to SEM (for example), where the apparent length of a feature can only be contracted (foreshortening [104]). For the SHeM, stretching is also possible.

Considering the geometry shown in Figure 3.13, one can use trigonometry to define a scaling factor (SF) responsible for transforming the true length of a feature into the apparent length in a SHeM micrograph. Defining θ as the angle between the feature in question and the base plane, and α as the angle of the beam relative to the base plane normal (see Figure 3.13), a linear feature in a SHeM micrograph will be distorted according to the relationship:

푆퐹 = 푠𝑖푛(휃) · 푡푎푛(훼) + 푐표푠(휃). (Eq. 3.3)

For the SHeM (α = 45o), the above equation simplifies to:

푆퐹 = 푠𝑖푛(휃) + 푐표푠(휃). (Eq. 3.4)

Figure 3.14 shows a plot of the SHeM scaling factor as a function of θ between 0º (where the face sits flat on the base plane) and 180º (as with 0º, but in the opposite direction). Between 0º and 90º (the latter being perpendicular to the base plane), the factor is greater than 1, leading to the projection being longer than that of the true geometry (Figure 3.13 a and b). Faces at exactly 90º will not see any change in length (Figure 3.13 c), while faces at an angle greater than 90º will appear shorter than reality ( |SF| < 1, Figure 3.13 d and e). A face parallel to the beam (i.e.: at 135º) will not present any visible area to the beam and hence be invisible, reflected by a scaling factor of 0. Also note that the negative values of the factor between 135º and 180º (Figure 3.13 e) indicate the projected face now lying in the opposite direction.

73 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.14 Plot of the scaling factor SF = sin(θ) + cos(θ) in a SHeM micrograph for a feature (needle) at an angle of θ (see inset) with respect to the sample slide. A plane tilted away from the incident beam (0º < θ < 90º) will appear longer in a SHeM micrograph, while those tilted towards the beam (90º < θ < 180º) will appear shorter. Negative values of the scaling factor (135º < θ < 180º) indicate the apparent length of the plane has reversed direction (for example, would appear to the right of the base of the needle as shown in Figure 3.13 e). Note that this scaling factor will only apply in the horizontal scan axis – in the vertical direction the image will be in direct correspondence with the sample.

To gain a better understanding of the effect of the scaling factor, a TEM grid was employed as a locator for small features by removing the central portion of the grid. The alterations were performed with a fine drill bit, the result of which was a series of bent spars both above and below the plane of the grid, presenting interesting imaging opportunities. Figure 3.15 shows an optical micrograph of the prepared sample stud, while Figure 3.16 shows matched optical and SHeM micrographs of a small area of the grid.

74 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.15 Optical image of the section of TEM grid to be imaged, mounted on the sample stud (white area) with carbon tape (black). A drill bit had been used to excise the central portion of the grid to allow it to function as a locator for other samples, resulting in damaged spars being bent at a variety of angles with respect to the plane of the grid. The box indicates the region imaged in Figure 3.16.

Figure 3.16 Scanning helium micrograph (left) and matched reflection optical microscope image (right) of a region of TEM grid mounted via carbon tape. The SHeM micrograph shows both masking and shadowing due to the carbon tape below the grid and the bent spars of the grid sitting above the bulk of the sample (respectively). Letters (a) through (f) correspond to features detailed in the text overleaf.

75 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

As with previously shown SHeM micrographs, the flat areas of carbon tape ((a) in Figure 3.16) appear quite bright under the helium beam (top left corner), with the intensity exceeding that of the copper grid surface. Evident along the left hand side of the micrograph (b) is masking due to the beam striking the sides of the grid, with the extent of the occlusion changing as the height difference between the carbon tape and grid increases towards the bottom. Indeed, the variation in the intensity and masking in the bottom half of the SHeM micrograph reveals details of the carbon tape shape lost in the optical micrograph (c) due to the limited depth of focus and colour balance. Some of the broken spars along the edge of the hole in the centre of the grid were bent upwards (d), and hence provide a feature above the bulk of the sample (including the rest of the grid). Under the helium beam, the bent section casts a strong shadow onto the grid and carbon tape below it (top centre of the micrograph). Furthermore, looking carefully at this shadowed region, a bright edge is visible against the underlying grid surface, belonging to the sides of the bent spars casting the shadow (e). The darkness of the hole in the top right (f) is due to the stainless steel of the sample stud directing little of the incident helium flux to the detector as well as the shadowing and masking provided by the bent spars. To help visualise the situation, Figure 3.17 shows a cross-section of the sample imaged in Figure 3.16 along with several instructive helium beam positions.

Figure 3.17 Schematic cross-section of the helium trajectories interacting with the bent TEM grid. Scan positions represented by (a) and (b) will result in dark areas in the final image (as in the top right of the micrograph) due to masking and shadowing respectively, with the bulk of the helium beam unable to make it to the detector. (c) and (d) will yield bright regions in the produced micrograph, with (c) the more intense as the mean plane is tilted towards the detector, as well as being closer to the entrance aperture.

76 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

The SHeM micrograph in Figure 3.16 has another surprising feature – the carbon tape appears brighter than the spars of the grid. While not readily apparent in the matched optical image, the grid sits directly on the tape near the top of the imaged region, with the tape dropping further away towards the bottom (as indicated by the change in masking). If we consider the uniform regions at the top, the tape exhibiting a higher intensity than the copper grid (despite the latter sitting closer to the detector) indicates a difference in sub-resolution topological contrast between the materials. As shown in Figure 3.18, TEM grid surfaces tend to be quite coarse as viewed optically, but a carbon tape surface will often exhibit a very smooth surface if laid down carefully (largely depending on the amount and quality of handling when assembling the sample). It was found in all subsequent SHeM micrographs collected with the prototype instrument that carbon tape appears quite bright under the helium beam. Traditionally comprised of a matrix material coated with a carbon filled acrylic adhesive, the carbon tape would tend to form smooth ‘glassy’ surfaces if not disturbed. The topic deserves further investigation through careful sample choice. One potential option is new applications of techniques such as UV lithography and reactive ion etching to precisely affect the morphology of a nanodiamond surface [105-107], thus allowing the production of carbon samples with known surface feature sizes and shapes which may then be compared under SHeM imaging.

Figure 3.18 Optical microscopy of a typical copper TEM grid such as that imaged first in the prototype SHeM. Detail callout shows the very rough surface of the grid, leading to diffuse scattering of the helium beam. Scale bar is 40 microns in length.

77 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

3.5. Experimental vs. Model Comparisons

The gas flow model used to help with the design phase had provided predictions for many of the key parameters for the prototype – for instance, centreline intensity of the beam, the count rate at the detector, and even some limited signal-to- background estimates. During the course of normal operation of the instrument, many of these predictions could be tested to evaluate the performance of the model. For instance, earlier in Section 3.1 the centreline intensity of the beam source was discussed and experimentally found to be (1.48 ± 0.10) x 1020 atoms / second / steradian with a 200 bar stagnation pressure. The gas flow model had predicted a value of 1.46 x 1020 atoms / second / steradian for the same conditions, well matched to the experimentally determined intensity.

One of the difficult parameters to investigate was the resolution of the instrument. The gas flow model makes assumptions with regards to the beam spot on the sample surface, namely that the divergence of the beam is so low that it will be the same size as the installed pinhole, and that there will be no shape to the beam profile. As such, the model simply predicts a 5 micron resolution with a 5 micron pinhole installed. Both of these assumptions needed to be examined in more detail, but this would have to wait until the second iteration of the instrument. With the limited time available, it was decided to attempt to determine instrument resolution by scanning over an edge and looking at how quickly the intensity changed. While not an ideal sample for what is traditionally termed a knife-edge scan, the TEM grids shown previously would suffice to establish an initial estimate.

With the problems with the scanning stages, extracting specific lengths from the produced micrographs was deemed prone to error unless special care was taken. In finding an edge profile for the purposes of determining the resolution, the vertical axis on the micrographs was much more reliable and as such the profiles were taken in this direction. Figure 3.19 below shows three such profiles through a section of an image taken of a TEM grid. For this particular micrograph, each pixel represents 1.7 microns (vertically), and the 7 second dwell per pixel consisted of two seconds wait and 5 seconds of collection. The longer than normal wait time was to allow for the detector stagnation volume to further equilibrate to reduce the noise, improving the reliability of the measurement. It should also be noted that before the profiles were taken, the image had a linear background subtracted in the vertical axis to remove the effects of a slight tilt to the sample slide.

78 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.19 Location of the three linescans across the edge of a TEM grid spar used to obtain an estimate for the resolution of the prototype microscope.

Averaging the three line scans results in the plot is shown in Figure 3.20. The error in the sample position is ±1 microns (on account of scan stage drift during the image acquisition), and the error in the signal was taken to be the RMS noise in the areas of the image corresponding to the TEM grid bars (some 110 counts). By measuring the distance required to bring the signal from 20% to 80%, an estimate for the resolution of the instrument may be found. The levels of 20% and 80% were chosen so as to avoid the noise at the top and bottom of the transition (as shown by the error bars). In similar experiments with other microscopies, a change between 25%-75% is more commonly used, but will yield a smaller resolution. The linescan indicates a resolution for the instrument of 3±2 microns, which interestingly is smaller than the spot size of the prototype (as the beam must diverge through the pinhole and hence be more than 5 microns in diameter). The result would mean either the pinhole had clogged and thus was producing a smaller beam spot - confirmed to not be the case by optical microscopy of the pinhole during experimentation - or that the shape of the beam profile was surprisingly sharp. Further analysis of the linescan is possible in order to derive the shape of the beam, but such work was not conducted until needed for the second iteration of the microscope (see Section 5.4). It should be noted that the resolution derived owes much to the strong contrast present in the image used. It was empirically found that the contrast between materials was a much tougher restriction on the resolution than minimum spot size.

79 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.20 Result of the averaging of three vertical line profiles across the edge of a TEM grid as imaged by the prototype SHeM. Note that the intensities have been normalised for ease of analysis. Looking at the sample travel required to move from 20% to 80% of the intensity, an instrument resolution of (3 ± 2) microns is obtained.

The count rate at the detector was one of the most critical parameters during the early planning of the instrument, as it would largely control if a feasible image could be generated. The model predicted ~2500 counts per second for a sample which scattered helium with a much broader distribution than would be expected for the majority of materials (so as to estimate instrument viability in the worst case scenario). The count rate found experimentally was also sample dependent, typically ranging from 1500 – 4000 counts per second for the tuned Hiden quadrupole. As such, the model was found to be in good agreement with experimental data. The range of values does however illustrate the dependency of some of the instrument parameters on what is being imaged – of particular note are the signal-to-background (S/B) ratio and the size of the noise present. Further complicating the latter is the rise time of the stagnation detector and the ability to simply collect for longer per pixel in order to improve image quality. The following estimates for micrograph signal-to-background ratios and noise levels are representative values indicative of commonly employed imaging conditions.

The background in an image is caused by helium not originating from the scattering of the beam with the sample position of interest entering the detector volume. Helium from the differential stage will leak into the sample chamber during beam operation, thus forming a stable background helium pressure along with the gas that scatters from the sample but does not enter the detector. Consequently, there

80 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

is a constant leak of non-signal helium into the detector volume which will contribute to the observed count rate. While the gas flow model included such leakages in order to produce a signal-to-background estimate, the realisation that the helium pressure in the differential stage gives rise to a secondary beam means that the assumption of a flat background term is false. Despite the broadness of the effusive distribution, the position and composition of the sample the secondary beam is incident on will now alter the amount of helium entering the detector not from the position of interest. Evaluating the size of the background is rather difficult in practise, as it would require a sample with a known scattering in order to subtract the counts which constitute real signal. Instead, we look for instances where the primary beam is blocked from the detector, leading to the collected count rate being (almost) entirely background.

The upper right hand corner of the SHeM micrograph shown in Figure 3.16 is an excellent example of such occlusion of the primary beam, yielding a large number of pixels with which to produce an estimate of the background. In the shadowed region of this micrograph, we see a count rate of 3250 counts/sec, while for the flat regions of carbon tape to its left we get 3750 counts/sec. Thus, the S/B ratio for this section of this particular micrograph is 0.15 (± 0.03, based on the RMS noise for both areas). However, looking at the brightest and darkest areas of the image and calculating the same ratio yields a value between 0.06 and 0.26. Such a range is consistent with the S/B ratios generated from other micrographs taken with the prototype and is a good representative metric. The S/B ratio generated from the gas flow model was as high as 1.63, with the discrepancy largely due to the effects of the secondary beam (as discussed in Section 3.2) raising the background count rate quite significantly.

The final major parameter to be investigated was the noise present in a collected image. In the gas flow model, it had been postulated that the noise would simply be shot noise (i.e. the square root of the signal) originating in the limited helium signal entering the detector stagnation volume. To avoid the potential for topological features below the resolution of the instrument affecting the recorded intensity, areas of pure shadow in the collected micrographs such as that mentioned above for Figure 3.16 were utilised. Here, the intensity per pixel averaged out to be 6495 counts with an RMS noise of ±80, and thus confirmed to be shot noise as predicted. The observation was verified with other SHeM micrographs, with all areas with similar areas of occlusion having an RMS noise within 5% of the expected shot noise.

81 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

As a point of comparison to other microscopies with regards to noise, we can make use of the Rose Criterion [108]. This states that in order to be visible, a feature must have an intensity that differs from the background by at least five times the noise level of the background. Taking the TEM grid spars from Figure 3.16 as the feature of interest and subtracting the background intensity as measured from the areas of occlusion, we arrive at a signal-to-noise ratio of approximately ten – well above the Rose Criterion. It should again be noted that such is value is representative of typical imaging conditions (with longer collection times driving the ratio up) Performing the similar analysis on other Mark I SHeM micrographs with appropriate background regions, signal-to-noise ratios between 7 and 19 were found.

A summary of the model predictions as compared to experimentally determined values is shown in Table 3.1 below. Overall, the model does an excellent job at replicating the behaviour of the instrument, excepting the case of the background levels due to the unaccounted for secondary beam.

Parameter Model Experiment

Centerline intensity 1.46 x 1020 (1.48 ± 0.1) x 1020 (atoms / sec / steradian)

Resolution (microns) 5.0 3 ± 2

Count Rate 2600 1500 - 4000 (counts / second)

Noise √푛 √푛 ± 5%

Signal-to-Background 1.63 0.15 ± 0.1

Table 3.1 Comparison of the modelled performance characteristics of the prototype microscope with those found experimentally for a 200 bar beam at 295K. Note that the count rate, noise, and signal-to-background are all dependent on the sample under investigation – the values given are representative.

82 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

3.6. Materials Studies

3.6.1. Fly Wing

The SHeM micrographs presented so far from the prototype have been of samples with well-known morphologies, investigated for the purposes of understanding image formation in the instrument and defining its abilities. It was also desired to demonstrate one of the chief benefits to the technique, namely the complete non- destructive nature of the probe-sample interaction. At the same time, imaging a surface with a higher degree of complexity would be an interesting test, especially something organic in nature. A wing from the common house fly (musca domestica) was chosen as the sample that would fulfil these criteria. Figure 3.21 shows the section of wing examined in matched optical and scanning helium microscopy, while Figure 3.22 displays the SHeM micrograph overlaid over a larger optical image to allow for a better understanding of the features visible in both. No sample preparation was performed before mounting the wing on a sample stud, achieved by gently pressing it into a section of carbon tape. The SHeM micrograph of the wing was produced using a 100 bar beam, and is 100 x 400 pixels in size. With a dwell time of 4 seconds per pixel, the entire image took just over two days to complete. Intensities peaked at just over 4100 counts per pixel, with the darkest shadows recording a value of approximately 60% of this maximum. No change to the wing was observed due to exposure to the neutral helium beam, even with the long imaging time.

Figure 3.21 Optical micrograph (top) and matched scanning helium micrograph (bottom) of a section of wing belonging to the fly species musca domestica. The SHeM micrograph reveals much more of the detail in the transparent surface, including the convoluted folds of the wing membrane. Note that due to the surface sensitive nature of the neutral helium probe particle, the SHeM micrograph contains only information from the top side of the wing (unlike the optical).

83 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.22 SHeM micrograph of a section of fly wing overlaid on a composite of optical images of the same sample. Such a large image (100 x 400 pixels) required the instrument to scan for approximately 48 hours. Scale bar is 300 microns in length.

84 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

If the wing was to be viewed under electron microscopy, it would require a conductive coating else the insulating surface would suffer charging and cause artefacts. In addition to the difficulties a fragile sample such as this would pose to an electron beam, the transparency of the membrane also proves a problem to optical microscopy. While more solid features such as the veins which support the wing and the hairs attached to it are clearly evident, the elaborate folds of the membrane itself (allowing the wing to change shape during flight) are difficult to resolve with any clarity. Furthermore, the wing membrane actually consists of two separate layers with individual features, meaning the top and bottom can be quite different. The optical image includes features from both sides, but discerning which side for each can be challenging.

Figure 3.23 Cropped section of the SHeM micrograph shown in Figure 3.21. Rescaling the colour based on the intensities in this smaller portion allows more of the detail to be visible, including the hairs attached to the top edge of the wing.

Immediately noticeable in the SHeM image is a much greater level of detail, with strong contrast observed due to the surface topology. The complex surface of the wing stands out much more vividly, in particular for the large folds towards the right hand side. The surface sensitive nature of the probe particle means that all the scattering occurs from the top side of the wing, and thus comparisons between the optical and SHeM micrographs reveals more information on the location of certain features. For example, one of the large veins in the optical on the left hand side does not appear in the SHeM image, meaning that it actually crosses the wing on the bottom side. The separation of the two layers which constitute the wing can also be seen along the leftmost edge where it slopes down to meet the carbon tape it is attached to. The mean plane variation across the surface of the wing causes the varying levels of intensity, and if the change from left to right is considered the overall curve of the wing can be seen. The finer details in shape lead to local

85 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

variations, such as the ripples in the center, and folds to the right. The latter exhibit masking and shadowing to various degrees, allowing for an intuitive grasp of the height and angle of each particular fold. Note also that the left side of some of the folds shows a brighter highlight where the plane is angled towards the detector aperture. Further details, such as the hairs along the top edge of the wing (see Figure 3.23), can be seen quite clearly despite being close to the resolution limit of the instrument. In the case of the hairs this is due to their shadowing of the beam, allowing them to picked out against a bright background.

3.6.2. Tin Spheres on Carbon

In addition to using the TEM grids of a known size and shape, there exists a wide range of standards commercially available for the purpose of microscopy calibration. One of the most common are structures that repeat over many different length scales, allowing not only the resolution to checked at any level of zoom, but even more complex problems (such as astigmatisms in the optics) to be identified and corrected. Agar Scientific offers what they term a ‘Universal Resolution Specimen’ of this type in the form of tin spheres laid down on a carbon backing (part number: AGS1937). With sphere sizes ranging from 30 microns down to less than 5 nanometres, it was an excellent choice for early SHeM imaging for size calibration purposes, and a more intricate look at masking effects under the helium beam. The sample was also originally hoped to have potential for a first investigation of chemical contrast, but discussions with the supplier revealed that the manufacturing process would likely result in a variety of carbon filled oils and solvents coating both the spheres and the substrate. Figure 3.24 shows SEM and optical micrographs of the sample in question, while Figure 3.25 contains a pair of SHeM micrographs of the sample as imaged with the prototype. Note that the micrographs in Figure 3.25 have been rotated clockwise by 90 degrees relative to previous SHeM micrographs (meaning the beam enters from the bottom of the image) to aid the reader in identifying the spheres.

86 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

Figure 3.24 Top down SEM (left) and optical micrographs (right) of the ‘Universal Resolution Sample’ available from Agar Scientific (AGS1937). SEM image from the technical data provided by Agar Scientific [109]. The sample consists of tin spheres laid down on a mechanically polished carbon substrate, with spheres sizes ranging from approximately 30 microns in size down to less than 5 nanometres. In addition to resolution and masking effects, the sample is of interest to SHeM due to the difference in the two materials (visible as the contrast in the SEM micrograph).

Figure 3.25 SHeM micrographs of the tin sphere sample. In the left image, each pixel represents a step of 1.1 microns, while on the right 1.3 microns. Note that the beam enters from the bottom (90 degree rotation as compared to previous SHeM micrographs) to guide the eye.

Immediately apparent in the micrograph is the strong topological contrast provided by the raised features sitting on a flat substrate. Each tin particle causes a dark masked region where it occludes the beam; however, the lowest intensity features are shadowed regions on the beamward sides of the spheres. At these positions, the helium beam strikes the sphere surface and is largely directed back towards

87 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

the pinhole, meaning that the reflected helium is much less likely to strike another surface and hence cause multiple reflections which would contribute intensity. The masked areas appear brighter by comparison, as once the beam scatters of the carbon substrate some portion of it either enters the detector directly (due to the width of the aperture), or after multiple further scattering events. Multiple scatterings due to a sample plane tilted towards the substrate were also theorised to be responsible for the bright region towards the bottom of each sphere, but the limited resolution and imaging time meant that confirmation had to wait until the next iteration of the SHeM (see Chapter 5 for further discussion based on that instrument)

Figure 3.26 Section of SHeM micrograph showing a single tin sphere.

As discussed earlier, the angle of the sample surface relative to a flat plane will cause distortions in its apparent size in the collected micrograph. The first implication for the tin particles is quite clear in Figure 3.25, namely a stretching of the particle in the vertical direction. The amount of stretch will change as the angle of the surface plane relative to the substrate varies, meaning parts of the sphere will be extended more than other. If we consider the one dimensional path of the beam across a sphere surface (i.e.: a vertical line of pixels in the micrographs of the sample), using the convention defined in Figure 3.14 we note that the top of the sphere visible to the beam will appear more stretched than the bottom. The net result is that the majority of the tin particle as it appears under SHeM is the top half, helping to explain the sizes of the shadowed region and the bright bottom of the sphere. As the other axis does not experience any distortion due to the angled beam, the diameter of the spheres as measured from left to right in the micrographs presented in Figure 3.25 will yield the correct value. The biggest spheres in both micrographs can be measured out to be approximately 14 microns in diameter, well matched to the maximum size of 30 microns as quoted by the manufacturer. Judging what constitutes the smallest visible sphere is a more complex proposition,

88 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

but we can unambiguously observe a particle of only 3 pixels in diameter. This equates to a size of around 4 microns, a value which tallies well with the resolution of the instrument as found earlier.

3.6.3. Polymer-Bonded Explosive

Polymer-bonded explosives (PBXs) are formed by mixing a crystalline explosive powder together with a polymer binder to form a material exhibiting a range of beneficial qualities including good safety, high strength, and the ability to be shaped or even machined. The material is typically 80 – 95% explosive powder, with only a small amount of the polymer added [110]. While PBX is capable of being imaged with electron microscopy [110], care must be taken to avoid over- exposure to the energetic electron beam. For safety reasons, the explosive powder can be replaced with an inert simulant to elicit an analogous mechanical response, or the sample size kept very small to limit the danger. Furthermore, the insulating PBX must be coated with a conductive overlayer to remove charging effects, although modern environmental scanning electron microscopy (ESEM) is capable of bypassing this requirement. As such, PBX is an ideal candidate for demonstrating the delicate nature of neutral helium imaging along with the lack of any extraneous sample preparation to obtain an image.

Figure 3.27 Micrographs of a section of polymer bonded explosive as imaged via SHeM (left) and a polarised reflection optical microscope (right). The sample in question had been mechanically polished flat to within 50 nanometres, leading to the likely source of contrast in the image being topological contrast below the resolution limit of the instrument (i.e.: roughness). Pixels in the SHeM micrograph represent a sample movement of approximately 4 microns.

A sample of a prototype PBX under development was obtained after it had been polished flat to within 50 nanometres and imaged in the prototype microscope

89 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

(Figure 3.27). It should be noted that like the fly wing shown previously, the PBX sample received no sample preparation (aside from the polishing) before being mounted in the instrument, and no change to its composition or morphology was observed due to exposure to the neutral helium beam. Due to the high percentage of explosive powder relative to the matrix material, when PBX is polished to reveal the microstructure, the majority of the exposed surface is comprised of polished facets of the crystals [110]. This behaviour can clearly be seen in both the SHeM and optical micrographs in Figure 3.27. The polishing procedure has limited the size of any height differences, but masking is still evident through comparison of the left and right sides of any of the larger crystallites – for example, the large crystallite towards the center of the SHeM micrograph. Also visible in both micrographs are the grooves left by the polishing process (diagonal striations across the larger crystallites). With the knowledge of the morphology and composition of the sample, the source of contrast in the SHeM micrograph is almost certainly topological in nature, and not chemical. The amount of polymer in the sample is small, and confined to small veins between the larger crystals of explosive. In these same areas, smaller crystals in a wide range of sizes can be observed, meaning a much rougher surface than the larger polished facets. Such topological contrast below the resolution limit of the instrument will appear in the produced micrograph as a region of reduced intensity, excepting the unlikely circumstance where a large majority of smaller facets aim towards the detector. The SHeM micrograph, while not indicative of chemical contrast, does seem to show the technique is particularly sensitive to mean plane difference, especially when considering the striations left by the polishing process. Once again, considering the roots of the technique in helium atom scattering, such a predication is perhaps unsurprising.

3.7. Discussion and Future Work

Looking at the performance of the instrument as a whole, the prototype is considered a great success, with images of a surface created with neutral helium using currently available equipment. The generated micrographs were intuitive, demonstrated the primary advantages of the probe particle (including the surface sensitivity and non-destructive nature), and allowed the effects of the 45 degree geometry to be considered. While no evidence of contrast mechanisms beyond topological was found, it was apparent that the sample choice and operation of the instrument were not satisfactory for such studies. Furthermore, the gas flow model

90 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

of the apparatus constructed to aid in the initial design phase proved to be a reliable estimator of the critical parameters. Despite these successes, the prototype scanning helium microscope was as its name suggests – a proof-of-concept to allow for valuable insight into the specific requirements of a system built from the ground up. Accordingly, the normal operation of the prototype, along with the quality of the images generated, suggested a series of changes to the design to be considered for the successor instrument.

If the next version of the instrument is to produce higher quality images as well as attempt to investigate the more complex contrast mechanisms proposed by theory, the first issue to be addressed is the helium signal at the detector. While the 1500 – 4000 counts/second achieved for the prototype was sufficient to produce reasonable images, higher count rates are needed to improve the quality of the micrographs in terms of observed noise, signal-to-background, and number of pixels. The latter would be achieved due to the dwell time per pixel being reduced without sacrificing the size of the collected intensity. One of the most prevalent problems was the time taken to complete a scan as a dwell in excess of 5 seconds was typical, meaning images would take anywhere from hours to even days to finish. In addition to faster scan times, a higher count rate would also allow for a move to a smaller pinhole than the 5 micron one used in the prototype. The intensity required for a change in pinhole would be more drastic than simply speeding up the scans – the helium flux through the pinhole will vary on the square of its radius, and hence even a small reduction will drastically reduce the available intensity. It should be noted that expecting an improvement in resolution simply by reducing the size of the pinhole makes the assumption that this aperture is the limiting factor. With the observed resolution already smaller than the nominal diameter of the pinhole (see Section 3.5 above), it might be expected that the shape of the beam is a more critical parameter, and one not as directly influenced by the size of the final aperture. As a result, more work would need to be done with regard to the optics of the instrument before a move that would throw away a significant portion of the beam was made.

The primary factor in increasing the count rate will be in the design of the chambers constituting the microscope, much as was the case in the initial design phase. Changes to the supersonic free-jet beam source are unlikely to result in an improvement to its output (barring increasing the stagnation pressure beyond 200 bar, which greatly increases both the cost and danger of the compression), and a more modern quadrupole will only grant a small increase in sensitivity. Moving to

91 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

a solenoidal ion source would be ideal, but the price, complexity and availability make this more of a future goal as opposed to an instant design choice. Rather, more significant and cost effective gains can be made by cutting down the beam line further and thus reducing the distance over which the beam can diverge. Of the overall beam length, realistically only the skimmer to sample distance can be altered in a meaningful way. The nozzle to skimmer separation is set by the requirement to minimise both the backscattered helium and the differential stage pressure (as discussed with regards to the twin beams), and the sample to detector distance will be set by the working distance from pinhole to sample due to the 45 degree geometry. Furthermore, due to the detector operating in stagnation mode, once the helium enters the detector aperture in the pinhole plate it becomes a part of the stagnation volume and hence contributes to the recorded signal at that position. Consequently, bringing the distance from sample to detector down is not nearly as influential as keeping the stagnation volume low and thus reducing the rise time. Reducing the skimmer to sample distance of 143 mm, the vast majority of which is the length of the differential stage itself, is then the primary goal. Figure 3.28 shows a cross-sectional schematic of the differential stage from the prototype, with the red arrow (approximately 94 mm in length) demonstrating the excess which can minimised in a future design.

The need to shorten the beam path length complicates the second design goal of increasing the pumping available to the differential stage so as to reduce the size of the secondary beam. As shown above in Figure 3.28, the prototypes differential

92 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

stage was poorly pumped due to the constriction of using a 2.75” tube connection, along with the turbomolecular pump being situated a short distance away via a bellows connection. To better remove the excess gas, the turbo must be directly attached much closer to the pinhole plate. Reducing the beam path length and yet opening up the chamber to allow for a turbo means a fundamental redesign of the differential stage is required, likely along with parts of the source or sample chamber to allow for the new shape.

The detector chamber can also be modified in order to improve performance. While a sheath for the quadrupole separated from the sample chamber by a gate valve worked well, the pumping for the stagnation volume was found to be critical in terms of obtaining a good response. Too little pumping and the dwell time per pixel would need to be unreasonably long, and too much would mean the equilibrium pressure would not be sufficient to yield a useful intensity. The latter case was found experimentally - during initial testing of the microscope, a move from the bellows connection to the turbo to something more direct was attempted. Increasing the pumping from the estimated 2 L/s left the instrument without a workable count rate, and the bellows connection was quickly reinstated. Ideally, the pumping on the detector volume would be adjustable, allowing it to be set for the particulars of the scan to be conducted. Incorporating a variable constriction in a new sheath design, optimally while keeping the stagnation volume at least the same (if not smaller), is then a concern for a future design.

The final change to be addressed is quite obvious – closed loop control of the stages responsible for raster the sample underneath the beam. Even excepting the faulty Attocube unit responsible for the drift issues as an anomaly, the ability to image an area reliably then perform a higher resolution scan of a specific region or feature is critical to more sophisticated studies. Many other microscopies, through closed loop control and tools built into the user interface, allow the user to perform a low resolution scan, highlight the region of interest, and then execute a much higher detail scan. Such functionality is highly desired in a future instrument in order to speed up the time taken planning a micrograph and improve the reliability of the scanned areas.

Such concerns, along with a short list of improvements to be made in terms of the ease of use of the instrument, dictated the design process for the successor microscope detailed in the following chapter.

93 CHAPTER 3: The Mark I Prototype SHeM – Experimental Results

94 CHAPTER 4: The Mark II SHeM - Design and Performance

CHAPTER 4

THE MARK II SHEM - DESIGN AND PERFORMANCE

The work on the Mark I scanning helium microscope had not only confirmed that the technique was indeed feasible using the available technology, but that satisfactory images of sample surfaces could be produced. The development and operation of the prototype had highlighted a range of areas where improvements to the design could be made, and this knowledge formed the basis for a new iteration of the instrument to be located at Newcastle. The ‘Mark II’ was designed from the ground up through both gas flow modelling (as with the Mark I) and CAD drawings, with the primary goals being: (a) a higher helium signal at the detector, (b) minimising the effect of the secondary beam, and (c) providing a modular system to facilitate future developments as they occur. This chapter outlines the features of the new instrument, then compares the experimental performance to that of the Mark I. The improvements made allow for more than better quality micrographs – they facilitate new types of imaging, and for the first time an investigation into the contrast mechanisms beyond topological effects.

The design of the Mark II SHeM was completed at the University of Newcastle in collaboration with Matthew Barr (with some of the work performed also appearing in Barr’s thesis [32]). Actual construction of the instrument itself was completed with the aid of Barr, Joel Martens and Kirren Thompson. Data collection was achieved with Barr and Martens; all subsequent analysis performed by the author.

4.1. Design Principles

With the success of the scanning helium microscope prototype constructed at the Cavendish laboratories, plans were immediately made to build a second instrument. This ‘SHeM 2.0’, henceforth referred to as the ‘Mark II’, would not merely be a copy of the prior system - it would instead use the lessons learned

95 CHAPTER 4: The Mark II SHeM - Design and Performance

from the Mark I to improve the instrument performance as much as possible. Furthermore, it should ideally be modular in nature, meaning that parts could be swapped out as needed and the system would be capable of taking advantage of new technologies (such as replacing the quadrupole detector with a solenoidal ion source). It should also be noted that this new SHeM was envisioned as a much more friendly system to external users as a means to expand the sample set open to the technique.

The prototype instrument had utilised existing chambers at the Cavendish laboratories for its construction, and as such certain restrictions were set on the nature of the system. Being designed for the sole purpose of neutral helium microscopy, the Mark II would be able to eliminate some of these limitations – in particular, the length of the beamline could be reduced considerably in order to maximise the helium signal. As discussed previously, larger count rates would enable higher quality imaging, an enhancement to the possible resolution, and open the door to the less powerful contrast mechanisms. The pumping on the differential stage was to be increased significantly, thus reducing the strength of the effusive beam and its deleterious effects on the produced micrographs. Pumping was also a focus for the detector volume, where a means to control the pump rate would mean the ability to vary the rise time of the detector. In addition to the larger changes, a host of other small features were also to be considered: adding cooling to the beam source; closed loop control for the sample stages as well as a more accessible sample mount solution; quicker and easier sample exchanges, which along with the user interface for scan routines should make the system amenable to external users. With these goals in mind, the system was designed with the aid of an updated gas flow model and 3D CAD packages.

4.2. System Overview

To help the reader contextualise the discussion of the major changes, a brief overview of the final instrument is useful. A top-down cross-sectional schematic of the Mark II microscope is shown in Figure 4.1. A supersonic free-jet expansion of helium originates in the source chamber, with the source now capable of chilling the stagnating gas down to approximately 100K. The 100 micron Beam Dynamics skimmer responsible for selecting out the centreline intensity of the expansion is mounted to a detachable plate between the source chamber and a box chamber housing both the differential stage and the sample chamber (separated by an

96 CHAPTER 4: The Mark II SHeM - Design and Performance

internal wall). The helium beam passes through the differential stage to an opening in the internal wall where it is incident on the pinhole, once again a FIB milled aperture in a silicon nitride membrane. The result is a thin beam of helium striking the sample, mounted on three Attocube linear drives in the sample chamber. A portion of the helium reflected from the surface will pass through a second hole in the pinhole plate, connected to the detector chamber via a short section of tube internally welded so as to keep it isolated from the differential stage through which it passes. The quadrupole housed in the detector chamber operates in stagnation mode, allowing the pressure to equilibrate before an intensity from that position on the sample surface is recorded.

Figure 4.1 Schematic diagram of the SHeM II system. Using a 10 μm nozzle, a supersonic free-jet expansion of neutral helium is produced in the source chamber. A skimmer samples the centerline of the expansion, which then passes into the differential stage. The optics of the instrument consists of a silicon nitride membrane with a FIB milled pinhole mounted into a metal plate (see inset). The beam is incident on the back of this membrane, leaving a small spot of helium free to strike the sample. The sample itself is able to be rastered underneath the beam via three linear drives to facilitate imaging. A portion of the helium reflected from each point on the sample surface passes through a second aperture in the pinhole plate into the detector chamber, where the stagnation pressure is sampled to yield the intensity.

97 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.2 Photograph of the Mark II SHeM constructed at the University of Newcastle. Visible on top of the system frame is the source chamber to the left, attached to the sample chamber (door with the window in the center). Along the back wall can be seen the gas panel responsible for the compression and regulating the helium supply to 200 bar, fed into the top of the source chamber through the visible thin pipework. The black insulated gas lines are those bringing chilled nitrogen into the same source assembly.

98 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.3 Photograph of the Mark II SHeM displaying (from left to right) the Hiden quadrupole control box, the detector chamber and pumping connections, the box chamber housing both the sample and differential stages, and finally the source chamber including helium and nitrogen gas supply lines. Note that the pipework for the chilled nitrogen gas (braided cables) have not yet been insulated in this photo (as compared to Figure 4.2 previous).

99 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.4 Gas flow schematic for the Mark II scanning helium microscope. High vacuum stages are shown in black, rough vacuum in blue, high pressure components in red, and gas storage in green.

100 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3. Instrument Construction

The following section details the various elements of the design of the Mark II instrument. As the geometry is an evolution of the Mark I presented in Chapter 3, the discussion will focus more heavily on the changes made to the design.

4.3.1. Source Chamber

A cross-sectional schematic of the source chamber as viewed from the front is shown in Figure 4.5. The main body of the chamber was cylindrical, with the pumping port located at the base. The chamber was capped with a custom adapter flange with extra DN16CF ports (used primarily for electrical feedthroughs), as well as a port for the X-Y-Z UHV-Designs manipulator which controlled the nozzle position. The helium source itself was suspended vertically from a vacuum feedthrough flange located at the top of this manipulator. It should also be noted that the adapter flange could be spun 180 degrees, thus enabling the source assembly to be pulled further back to increase the length of the beam line (for reasons that will be discussed later in Chapter 5).

The source chamber was pumped by an Edwards STP-iXR2206 turbomolecular pump capable of 2200 L/s on an ISO250 inlet flange, backed by an Edwards E2M80 rotary vane pump situated in an adjoining room for reasons of sound. The chamber had an ultimate pressure of 5 x 10-9 mBar, which during operation of the beam source (at 200 bar stagnation pressure) typically rose to a pressure of ~1.5 x 10-3 mBar.

Between the turbo pump and the chamber itself was an ISO250 pneumatic gate valve which allowed the pump to be isolated from the source chamber if required (for instance, in the event of a power failure during beam operation). The gate valve was sat within a cart which ran on 6 linear bearings linked to a guide rail attached to the frame, thus allowing the entire source chamber (once detached from the rest of the system) to be slid backwards. Movement of the source chamber in this way enabled excellent access to both the skimmer and the nozzle assembly for maintenance or alterations when needed.

101 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.5 Cross-sectional schematic of the Mark II SHeM source chamber as viewed from the front. The source chamber (green) was pumped by a large turbo pump (Edwards STP-iXR2206) through the bottom ISO250 flange. A nozzle assembly (red) based on the style of Buckland et. al. [55] with a novel cooling system was mounted to an UHV Designs manipulator (blue) allowing precise control of the nozzle relative to the skimmer.

102 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.2. Helium Beam Source

The performance of the supersonic free-jet source, utilising a 10 micron aperture, was evaluated in much the same way as with the Mark I SHeM. A plot of the corrected source chamber pressure was made as a function of stagnation pressures from 20 to 220 bar at room temperature (Figure 4.6). The data was well approximated by a linear fit (R2 value of 0.991), the slope of which allowed a determination of the effective nozzle diameter of 10.10 ± 0.37 microns using equation 3.1. With a 200 bar stagnation pressure, the helium partial pressure in the source chamber as measured by an ion gauge was (6.72 ± 0.2) x 10-3 mBar. Using equation 3.2 with the chamber pump rate of 1.5 m3/sec and a temperature of 298K, we calculate a centreline intensity of (1.56 ± 0.05) x 1020 atoms / sec / steradian for the Mark II SHeM source (very similar to that of the Mark I instrument).

Figure 4.6 Plot of the corrected source chamber pressure as a function of beam stagnation pressure for the Mark II microscope. Linear fit to data has an R2 value of 0.991 (indicating the source was functioning correctly) and a slope of (3.39 ± 0.25) x 10-5. From the latter, we derive an estimate for the effective nozzle diameter of 10.10 ± 0.37 microns, in good agreement with the nominal diameter of 10 microns.

The supply of helium was controlled via what was termed the gas panel, visible behind the system in the photograph in Figure 4.2. Bottled helium is sent to a gas booster placed in an adjacent room to be compressed to 200 bar, then subsequent two stage regulation ensures a constant stagnation pressure. It should also be noted the helium is passed through a cold trap before entering the vacuum system in order to remove any impurities.

103 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.3. Nozzle Cooling

With the greater focus on contrast mechanisms beyond topological, a means to control the energy of the helium atoms incident on the sample surface (and hence the dynamic surface processes they may interact with) needed to be added to the new microscope design. As shown in Figure 4.7, the solution was a novel approach to the cooling of the stagnation volume, one which also acted as the support structure for the vertically mounted source. A stainless steel pipe, fixed to a vacuum flange feedthrough, supports the main body of the source. Smaller diameter pipework is used to supply the compressed helium gas to the Buckland style nozzle assembly, around which a larger copper pipe was fitted. The result is a volume around the outside of the helium feed line, isolated from both the helium and the outside vacuum, through which chilled nitrogen gas could flow. Using the support pipe as the nitrogen inlet, a counterflow of nitrogen could be established, effectively cooling the stagnation volume.

Figure 4.7 Schematic cross-sectional diagram of the helium beam source for the Mark II SHeM design. A Buckland-style nozzle assembly [55] with a (nominally) 10 micron aperture was constructed, capable of withstanding stagnation pressures of at least 250 bar at cryogenic temperatures. The stainless steel pipework that supplied the compressed helium to the nozzle was passed through a larger outer pipe as shown, allowing a counter current of chilled nitrogen gas to provide cooling. Copper blocks on the nitrogen inlet and surrounding the nozzle assembly, linked by a copper bridge piece, acted to stabilise the temperature. Surrounding the nozzle copper block was a heater clamp, allowing the stagnation temperature to be precisely set via a PID controller.

104 CHAPTER 4: The Mark II SHeM - Design and Performance

Copper blocks were attached to the nitrogen inlet pipe, as well as the VCR cap containing the nozzle assembly, to regulate the temperature. Around the latter, a heater clamp (RS part number XQ-009-RS, customised to suit a high vacuum environment) was fitted so the nozzle temperature could be controlled. The current supplied to the heating element was regulated by a Eurotherm 3207 PID controller in conjunction with a K-type thermocouple embedded in the nozzle assembly copper block, allowing for closed loop control over the temperature and thus the beam energy. To ensure uniform behaviour, the two copper heat sinks are linked via a copper bridge piece.

Figure 4.8 Photograph of the source assembly for the Mark II. Surrounding the hexagonal VCR cap at the centre (holding the 10 micron aperture) is a copper block followed by a heater clamp to enable control over the temperature of the stagnation volume when used in concert with the counterflow of chilled nitrogen gas. Also visible to the top left is the secondary copper block on the nitrogen inlet pipe, as well as copper braid connecting the two cooling blocks (an initial linkage that was eventually replaced by the solid copper bridge shown in Figure 4.7).

From the gas panel, the nitrogen used to chill the assembly was regulated with a 10-100 litre/minute flow control meter (Dwyer VFA-27-SSV-PF). Exiting the panel, the nitrogen passed through a copper coil sat within a modified 12 litre dewar which was filled with liquid nitrogen. The result was cold gaseous nitrogen (along with a small percentage of condensed liquid nitrogen once the temperature of the lines dropped sufficiently) which was then forced by the gas pressure up through flexible pipework to the feedthroughs supplying the source assembly within the chamber.

105 CHAPTER 4: The Mark II SHeM - Design and Performance

The performance of the cooling system was characterised by use of K-type thermocouples (read via a Fluke 52 II thermocouple thermometer) attached to the nitrogen pipework at the top of the source chamber manipulator and the nitrogen inlet copper block in the source chamber, as well as the thermocouple embedded in the nozzle copper block as monitored by the Eurotherm PID controller. A plot of the temperature data of the three monitoring positions, as well as the source chamber pressure, is shown in Figure 4.9. Stable temperatures can be seen to be reached after 80 minutes, with the copper heat sink at the nozzle taking the longest to settle to a final temperature of approximately 130K. The difference between the nozzle heat sink and the one attached to the nitrogen inlet (minimum of 95K) can be attributed to the difference in thermal contact with the chilled nitrogen flow, and is also the likely cause of the delay in the former settling to a final temperature. The difference in temperature between the various copper blocks led to a search for another means with which to verify the temperature of the produced helium beam. Such a requirement has been previously noted with other chilled supersonic beam sources where the temperature of the nozzle itself does not necessarily translate to the final gas temperature [111- 113]. The difference is thought to arise from the dissimilar thermal conductivities of the stainless steel components and the helium gas, as well as the temperature gradient along the mount and radiative heating from the chamber walls.

Figure 4.9 Plot of the temperatures at the top of the source chamber manipulator as well as the two copper blocks as a function of time during operation of the cooling system. Also included is the source chamber pressure, observed to increase due to the change in centreline intensity predicted by equation 3.1 with gas temperature.

106 CHAPTER 4: The Mark II SHeM - Design and Performance

Knowledge of the centreline intensity given the chamber pressure (equation 3.2), and a relationship between that intensity and the stagnation temperature (equation 3.1) means that the source chamber pressure can be scaled to estimate the temperature of the gas. The final chamber pressure in Figure 4.9 of (1.22 ± 0.1) x 10-2 mBar can then be found to translate to a temperature of 97 ± 3 K, much lower than the nozzle temperature would indicate. With the knowledge of the performance of the cooling system, gas temperatures were henceforth determined through careful observation of the source chamber pressure during beam operation.

4.3.4. Skimmer Placement

In order for the length of the beamline to be flexible, it was necessary to move away from mounting the skimmer on a dedicated vacuum flange. The basic mounting design was however quite successful, and so was replicated with the skimmer being clamped down to a mounting plate located between the source and differential chambers. In particular, the Mark II SHeM employed a 100 micron ‘Model 2’ skimmer as produced by Beam Dynamics Incorporated [114]. The shape of the mounting plate used yields control over the skimmer position, and hence the distance the helium must travel to the sample surface.

Figure 4.10 3D render of the skimmer mounting plate (blue) and clamping ring (red) used to secure a 100 micron ‘Model 2’ skimmer as produced by Beam Dynamics Incorporated. The mounting plate may be swapped out for other versions with different depths, thus allowing the distance from skimmer to pinhole to be varied.

107 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.5. Differential Pumping

The alterations made to the differential pumping stage is among the most significant changes made to the Mark I design. A cross-sectional schematic of the differential pumping stage as viewed from above is shown in Figure 4.11.

Figure 4.11 Top-down cross-sectional schematic of the redesigned differential pumping stage. On the left, the 8” CF flange forming the edge of the source chamber (including nozzle assembly in red) is visible. The flange joins to the side of the box chamber (brown) containing both the differential and sample chambers, separated by an internal wall. Just inside the connection, the skimmer mounting plate (see Figure 4.10) sits, aligning the skimmer to the beam axis and setting the skimmer to sample distance. The center of the expansion produced by the nozzle is selected out by the skimmer, passes through the differential stage, and enters the back of the pinhole plate via a hole in the internal wall to be incident on the silicon nitride membrane with pinhole. Reflected helium from the sample is able to pass through to the detector volume via a 1 1/3” half nipple internally welded to ensure isolation from the differential volume. Note the visible 4.5” port on the bottom of the differential stage, allowing the direct connection of an Edwards EXT75 DX turbomolecular pump to ensure maximum pumping to the volume.

108 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.12 3D render of the connection between the source and differential chambers, with various levels of sectioning to help illustrate the complex geometry used to ensure a condensed beam path while still providing significant pumping.

From the Mark I instrument, it was apparent that the configuration of the differential stage was hampering the available pumping, leading to the formation of the secondary effusive beam. In order to improve the pumping while working to reduce the beam path length, the shape of the chamber had to be reworked from scratch, with the final result being that the differential and sample stages were incorporated into a single box chamber separated by an internal wall. The change allowed the extent of the differential stage in the direction of the beam to become quite short (thus improving the count rate), while broadening at the sides enough to allow for a pump on a 4.5” port to be attached to the volume directly. In particular, an Edwards EXT75 DX turbomolecular pump was used to evacuate the stage, backed by the same E2M80 rotary pump as the source chamber. The ultimate pressure for the chamber was better than 5 x 10-10 mBar with the beam not in use. It should also be noted that the ion gauge measuring the pressure was located at the top of the chamber, almost directly above the exit from the skimmer mount plate. As such, its readings are much more representative of the helium partial pressure than that in the prototype instrument.

109 CHAPTER 4: The Mark II SHeM - Design and Performance

As shown in Figure 4.11 and Figure 4.12, the skimmer mounting plate was attached just inside the 8” flange connection between the source and differential stages. While the back of the plate comes close to the internal wall separating the differential and sample volumes, the open space around the edges ensures pumping on the back of the skimmer is not significantly impacted. The centreline of the beam selected out by the skimmer is able to pass through a gap in the internal wall and enter the pinhole plate, bolted to the wall from the sample side. The shape of the chamber makes it then necessary for the helium reflected from the sample to traverse back across the differential stage. The DN16CF half nipple connecting the sample chamber to the detector extends into the box chamber and across the differential stage, internally welded so as to keep the two volumes separate.

As previously mentioned, access to the skimmer and the internal volume of the differential stage was granted by sliding the source chamber back on linear bearings. A bypass between the source and differential chamber volumes was also added to prevent the possibility of damage to the skimmer during pump down procedures, or in the event of a power failure. The reduction to the length of the beam path and the effect of the improved pumping on the size of the effusive beam will be discussed further in Section 4.5.

4.3.6. Sample Chamber

The move to a box chamber to contain the differential stage also meant a large change to the shape of the sample chamber. When considering the design, it was important that as the instrument would be used frequently for a large variety of samples, sample interchange should be a quick and easy process. Modern commercial microscopes utilising vacuum chambers often employ a quick release door to access the sample, a feature echoed in the Mark II SHeM. The entire front wall of the chamber (visible in the photograph in Figure 4.2) was made into hinged door sealed via a viton gasket, with an 8” viewport set into the center to allow a view of the sample stages within.

The sample chamber is connected to a 520 L/sec Pfeiffer TMU 521P turbomolecular pump via an 8” port on the right hand wall of the chamber, backed by an Edwards RV5 rotary vane pump. The result is an ultimate pressure of approximately 1x10-8 mBar, depending on the choice of sample. More importantly, the process of swapping out the sample and bringing the system back down to an

110 CHAPTER 4: The Mark II SHeM - Design and Performance

acceptable pressure for scanning (deemed to be anything less than 1x10-6 mBar) could be completed in an hour, a significant reduction on the Mark I system. A bypass between the sample and differential chambers in the backing lines allowed for the equalisation of pressures when bringing the chambers up or down from atmosphere to minimise stress to the pinhole membrane.

As discussed, the internal wall plays host to the pinhole plate containing the silicon nitride membrane which forms the final optical element for the helium beam. The pinhole plate itself remains largely unchanged from what appeared in the Mark I instrument, except for the removal of the mounting points for the sample mechanism. Viton gaskets are set into grooves in the back of the plate around both the beam entry and the detector exit, acting to seal the sample chamber from both the differential and detector volumes when the plate is fastened to the wall. The gaskets minimise the helium leakage from chamber to chamber, thus lowering the background levels present for any micrograph collected.

Figure 4.13 Top-down cross-sectional schematic of the box chamber comprising the differential and sample volumes. Note that for simplicity, the sample mount and the hinged door which forms the front wall have been omitted.

111 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.7. Sample Mount

The sample mount utilised in the prior version of the instrument, based on that of the scanning transmission x-ray microscopy (STXM) endstation at the Advanced Light Source, was found to be quite robust in terms of ease of use and reliability. As such, a similar mount with a few improvements was employed in the Mark II, a 3D render of which is shown in Figure 4.14. The mount is made up of a base plate, a stand comprised of three support rods and a platform on which rest the three 3 linear drives to provide sample movement. Finally, on the face of the third linear stage a kinematic mount is attached into which trapezoidal sample slides are able to be placed. The shape of the sample slide and placement of the phosphor-bronze pins in the kinematic mount forces the slide to naturally lock in to the same position in the Y axis, while also preventing movement in the XZ plane. The base plate of the assembly was bolted to the floor of the sample chamber in front of the pinhole plate as can be seen in Figure 4.15

Figure 4.14 3D render of the full sample mount which consists of a base plate, stand, three Attocube ECS3030 units to drive the sample in the X, Y, and Z axes as shown, and the three-pin kinematic mount into which the sample slides are placed. Inset shows a photograph of the stacked ECS3030 stages and kinematic mount with sample slide.

112 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.15 Photograph of the inside of the sample chamber with the top level of the sample mount removed, illustrating the placement in front of the pinhole plate.

The primary weakness of the mounting scheme in the Mark I instrument was the position of the sample in the Z axis being handled by a screw within the sample slide itself. Setting the face of the sample to the specular position was time consuming, and the design allowed no variation of the Z-position without removing the slide from vacuum. Furthermore, the screw unbalanced the weight distribution, meaning that if not placed carefully the slide would slip forward within the guiding pins, causing a tilt to the sample surface. Avoiding these undesirable behaviours motivated the move to linear drives in all three axes, permitting the sample not only to be rastered in front of the beam, but to be moved in and out of the specular position as desired. An unintended consequence of the enhanced Z control was the ability to vary the effective angle subtended by the detector aperture, opening the potential to stereoscopic imaging with the instrument (as discussed in Section 5.1.1).

The linear stages are once again Attocube ECS3030 slip-stick piezo drives in a UHV compatible configuration; however the new units were the closed loop variants (ECS3030/NUM) with position encoders. In conjunction with the ECC100 controller unit (interfaced to a custom built LabView program), the new stages were able to be driven with a quoted precision of 1 nm [115] and maintain an absolute

113 CHAPTER 4: The Mark II SHeM - Design and Performance

position reference at all times. The move to the encoded stages eliminated the problems with drift common to the previous instrument. With the stability provided by the kinematic mount and the closed loop control, it was found that inserting previously imaged sample slides returned features to within 50 microns of their prior positions.

The other issue of easy access to the sample slide was solved in making the entire front of the sample chamber a viton sealed door, with adequate space above the mount to allow the user to simply reach in, pull out the previous slide, and drop in a new one. In aligning the sample to the beam axis, the sample chamber has a DN16CF half-nipple with glass viewport located on the top of the box chamber, directly above the sample location. Through the use of a webcam connected to the control PC, driving the sample to the specular position could be performed while monitoring the separation between the pinhole plate and sample slide directly.

4.3.8. Detector

In designing the next SHeM, finding a highly sensitive commercial quadrupole was a strong priority. Based on the performance of the Hiden Analytical quad used in the Mark I system, it was known that whatever was to take its place needed to (at the very least) match its sensitivity of 3 x 10-5 A/mBar. Also necessary was the ability to connect to an external pulse counter (Agilent 53230A) and thus allow for the intensity per pixel to be recorded during scans. Finally, the physical size of the quadrupole head was a concern, as keeping the stagnation volume low would mean for a quick response in terms of the rise time of the instrument. The quadrupole which ended up satisfying all of these requirements was the updated version of that used in the Mark I instrument, the Hiden Analytical HAL/3F-PIC 50 AMU residual gas analyser (RGA). The more recent iteration has a quoted sensitivity of 1 x 10-4 A/mBar for helium-4.

With the choice of detector in the Mark II SHeM made, and thus the dimensions of the quadrupole probe known, attention was turned towards the design of the chamber into which it would be placed. A sectioned 3D render of the final detector chamber can be seen in Figure 4.16, along with the quadrupole probe and RF head for the Hiden HAL 3F-PIC. The main chamber consists of a sheath tailored to the shape of the quadrupole probe, with a pneumatic DN16CF gate valve incorporated into the sample chamber end. The valve allows the detector volume to remain at a good ultimate pressure by isolating it from the sample chamber during sample

114 CHAPTER 4: The Mark II SHeM - Design and Performance

interchanges. Two DN40CF ports are built into the bottom of the sheath at either end to provide pumping connections, with the size of the constrictions controlled by a pair of butterfly valves. The pumping for the detector is provided by a pair of Edwards EXT75 DX turbos in series (for better compression and hence lower ultimate pressure), backed by an Edwards RV5 rotary. The turbos are connected as directly as possible to the two pump ports through a custom DN40CF cross in order to maximise the available pump speed.

The two main priorities for the section, based on the experience with the previous system, were to minimise the volume around the quadrupole in order to improve the minimum response time for the stagnation, and to provide tunable pumping so that the response could also be varied as desired. The level of control and range required over the latter was not yet established and so butterfly valves were chosen for easy, if coarse, control over the throughput. Allowing the detector volume to potentially be pumped from both ends was a more practical choice. With the sheath being very close to the sides of the quadrupole, keeping the chamber clean (and thus maintaining a good ultimate pressure) would potentially take a significant amount of time if only pumped from the front. In practise, when a scan was being performed the back butterfly valve was fully closed and the front one varied as desired. Opening the back valve at the end of a scan quickly brought the chamber down to its ultimate pressure of 4 x 10-9 mBar.

Minimising the total stagnation volume required keeping the distance between the pinhole plate and the ioniser of the quadrupole as small as possible. The main sheath surrounding the probe was kept tight, resulting in a volume for this section virtually identical to that of the Mark I design, despite the addition of the two pumping ports. Note that the stagnation volume can be considered to end at the butterfly valve faces, based on the almost direct connection between the turbo pump and valves. As such, removing any extraneous length between the back of the pinhole plate and the ioniser would result in a smaller stagnation volume. In the Mark II, the gate valve used to isolate the sample chamber from the detector during a sample interchange was incorporated directly into the main body of the chamber. Doing so removed the need to leave enough length for the bolts required to connect the flanges together, and hence represented a significant reduction in volume. This change, along with the removal of the bellows connection to the turbo pump (which represented an addition to the stagnation volume for the Mark I design), meant that the stagnation volume for the second generation system was 40% smaller than that previously.

115 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.16 Sectioned 3D drawing of the detector chamber. The Hiden quadrupole (red) sits tightly within the main body of the chamber, minimising the stagnation volume. Further in this regard, the DN16CF gate valve (green) used to isolate the detector has been incorporated directly into the sheath. A small turbomolecular pump (Edwards EXT75 DX) is connected to the main body of the chamber via two DN40CF butterfly valves (blue). By changing the extent to which each valve is opened, the pumping on the stagnation volume can be controlled by the user.

116 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.9. Frame and Vibration Isolation

The frame on which the vacuum chambers were supported was constructed in much the same way as that for the prototype; heavy steel I-beams forming a base frame on which was sat a sub-frame built from hollow square section. The footprint of the frame was kept fairly compact (approximately 1.5 m square), but more tabletop was added than required with the view to eventually replacing the compact detector based on a commercial quadrupole with a much more powerful solenoidal ion source. The entire frame was held off the ground by means of three custom built vibration isolators placed equidistantly at the three corners of the base frame. Each isolator consisted of an airbag with which to damp motion in the vertical axis, and soft rubber to remove vibrations in the horizontal plane. As each airbag is capable of supporting 1.5 tonnes of weight independently, the frame will quite easily tolerate any and all future changes to the hardware of the instrument. The space under frame was also designed to keep the various electrical, water, data and gas supply utilities well out of the way, as well as to store the smaller rotary pumps needed as backing for several of the vacuum chambers.

Figure 4.17 3D render of the system frame, constructed from steel I-beams (red) and hollow square section (green). The entire frame was supported by three novel vibration isolators (callout). Also visible are the linear bearing rails (yellow) used to support the source chamber cart and allows access to the nozzle assembly or skimmer as required.

117 CHAPTER 4: The Mark II SHeM - Design and Performance

4.3.10. System Control and User Interface

As mentioned in Chapter 2, the Mark I instrument’s scan routines comprised of a series of Matlab scripts. However, moving the sample to the starting position of a scan was done via the control box for the Attocube units, a process that (with the lack of closed loop control) was time consuming and inaccurate. For the Mark II SHeM, a full LabView scan interface with control over the Attocube stages was constructed to allow functionality similar to that of any other modern microscopy. A screenshot of the interface can be seen below in Figure 4.18.

Figure 4.18 Screenshot of the LabView user interface for the Mark II SHeM. In addition to the basic control over the current position of the sample and the parameters of the scan to be run, the program also shows the image currently being generated along with a profile for the current horizontal line. Furthermore, once a scan is complete, the datafile can be reopened in this interface and a region boxed to start a new scan. As such, imaging features of interest within a prior scan becomes very quick and easy, especially as compared to the Mark I instrument.

The LabView interface contains the controls for steering the sample in all three axes, along with a readout of their current position as indicated by the internal interferometers, allowing for precise and repeatable positioning. Raster scan parameters can be edited, including the scan size, step size, wait and collection times, and the progress of the current scan is displayed. Furthermore, previous scans could be opened in the interface, a graphical area selected, and then this

118 CHAPTER 4: The Mark II SHeM - Design and Performance

smaller section scanned to allow for quick investigation of features of interest. In addition to raster scans, a second scan mode allowed line scans to be completed at a series of Z distances with automatic compensation built in for the 45 degree geometry (so the feature of interest always remained in the desired position relative to the beam).

An interesting consequence of the updated interface (as well as the higher count rate) was a new scan procedure became routine for the Mark II instrument. When a sample slide was first loaded in, a large area ‘locator scan’ with low pixel count was conducted to give the major features. After optionally matching the SHeM micrograph to an optical microscope image, regions of interest could be selected for higher resolution scans, drastically reducing the setup time for each new sample.

Figure 4.19 Illustration of the ‘locator scan’ used when first loading a new sample slide into the Mark II SHeM. An optical image (typically either an optical micrograph or, as in the left image above, a photograph) would be matched to a quick, large area SHeM scan (right image) in order to identify all major features. By selecting smaller regions of this large scan, high resolution images were produced. Scale bar is 1 mm in size.

4.4. Path Length

The most significant improvement to the design over the previous generation of instrument is a reduction in the total length of the beamline. While a longer beamline is traditional in helium atom scatterers in order to improve the angular resolution and allow for the differential pumping required to lower the background to acceptable levels, it also reduces the intensity reaching the sample. To qualify where the reductions to the path length were achieved, we consider breaking the total length into smaller sections. As illustrated in Figure 4.20, we will define

119 CHAPTER 4: The Mark II SHeM - Design and Performance

distances between the nozzle and skimmer aperture (ZNS), the skimmer and the

pinhole (ZSP), the pinhole and sample (ZWD or working distance as noted

previously), and finally the sample and detector (ZSD). Note that the latter measurement is taken as the distance from the sample to the front of the quadrupole ioniser, and is quoted to illustrate the change in stagnation volume (as discussed in 4.3.8). Table 4.1 compares the beam path length using this nomenclature between the Mark I and Mark II instruments.

Figure 4.20 Helium beam path through the microscope geometry, broken down into smaller sections for the purpose of comparing the improvement of the Mark II over its predecessor.

Distance Mark I Mark II

ZNS 10 mm 11 mm

ZSP 140 mm 49 mm

ZWD 3.0 mm 2.8 mm

ZSD 222 mm 180 mm

Table 4.1 Comparison of the distances making up the total beam path length for the two SHeM systems. The changes to the shape of the chambers mean the Mark II beam path length is much shorter, contributing to an improved count rate.

120 CHAPTER 4: The Mark II SHeM - Design and Performance

The nozzle-to-skimmer separation is determined by considerations of the reflections from the skimmer (acting to attenuate the central intensity of the beam) and the amount of helium passed into the differential pumping stage. The latter concern is less imperative than in the prior system – as will be discussed shortly, even with the nozzle at closest approach (namely 6 ± 1 mm) the supersonic beam is seen to easily dominate the effusive. Evaluating equation 3.2 for the experimentally determined parameters for the Mark II instrument (a nozzle diameter of 10.1 um, a 200 bar stagnation pressure at a temperature of 298K, and a source chamber helium partial pressure of 6.72 x 10-3 mBar) yields an ideal separation of 10.1 - 12.1 mm, depending on the choice of c. The majority of the imaging performed with the system was then done with ZNS set to 11 mm.

The largest change evident in Table 4.1 is the reduction in ZSP, made possible by the alteration to the differential chamber geometry (while still allowing an improvement in the pumping). Many small modifications to the design contributed to the this reduction; for instance, pushing the skimmer further into the differential volume via its mounting plate, reducing the thickness of the internal wall, and even thinning out the pinhole plate. The latter change is also responsible for the slight reduction in working distance, but the modular nature of the design means that this separation can quickly be reduced further. Extending the length of the cone of the pinhole plate drops the working distance and drastically raises the count rate, but the increase comes at the expense of the angular resolution and the potential for poor pumping around the sample. The change in working distance meant an increase (as compared to the Mk I instrument) in the angle subtended by a 5 micron pinhole (2.4 x 10-6 sr) and the 1 mm detector aperture (9.39 x 10-2 sr) at the nominal imaging position.

Indeed, all of the microscope components critical to the source to sample distance – nozzle, skimmer and pinhole – are all intentionally mounted on easily exchanged components to permit further alterations. It is estimated that as long as careful considerations of potential obstructions to local pumping are made, an additional 5 – 15 mm could be removed from the skimmer to sample distance.

121 CHAPTER 4: The Mark II SHeM - Design and Performance

4.5. Critical Performance

The most fundamental changes to the instrument between the two iterations were those designed to address the major issues of the Mark I: skew artefacts in the produced images, a higher helium signal at the detector, and the strength of the secondary effusive beam. The first problem was solved through the move to encoded linear drives in all three axes. With accurate readouts of the sample position and the ability to directly control the distance travelled between pixels, no drift was observed in any of the micrographs produced by the Mark II SHeM. Furthermore, a script was written to pull the scan parameters from each datafile and output a Gwyddion file (an open source scanning probe microscopy analysis suite) with correctly scaled X and Y axes. The software allowed for easy comparisons between SHeM micrograph and those taken with optical, AFM, or SEM systems. The SHeM micrographs (such as that shown in Figure 4.22 below) were found to reproduce the features and distances obtained with other microscopies with good agreement.

Figure 4.22 Matched optical (left) and neutral helium (right) micrographs of a copper TEM mounted to a sample slide with a central hole. SHeM micrograph produced using a 1.3 second dwell per pixel and a step size of 12.5 um. Scale bar is 200 microns in length.

The final count rate of the instrument was estimated throughout the design process by an updated version of the gas flow model used in the planning for the Mark I instrument. While not able to reproduce the effects of the effusive beam, it had been shown to be an accurate predictor of the helium signal entering stagnation volume. With the new instrument geometry inserted into the program, a 200 bar beam at a temperature of 295K was predicted to have a count rate of the order of 35 000 counts per second. Such an improvement over the 2 600 counts per second output for the Mark I system was due to the reduced beam path length and the

122 CHAPTER 4: The Mark II SHeM - Design and Performance

higher sensitivity detector (approximately a 5 and 3 times improvement from each, respectively). The experimentally determined value for the Mark II was found to agree well with the estimate, with a count rate typically between 20 000 and 40 000 counts per second, depending on the sample under investigation and the size of the pumping on the stagnation volume. The order of magnitude improvement in count rate over the previous generation of instrument meant that dwell times per pixel could be reduced significantly, making image collection much faster. Furthermore, the higher signal would open the door for subtle contrast mechanisms (inelastic effects) to be utilised for imaging, as discussed in the following chapter.

Figure 4.23 Detected helium signal as the nozzle is scanned across the skimmer for a series of nozzle-to-skimmer separations for the Mark II SHeM. For said experiment, a 200 bar beam at room temperature was utilised. In stark contrast to the similar plot collected for the Mark I, even with the nozzle at closest approach the free-jet beam is observed to dominate, although the presence of the effusive beam is still visible in the broad shoulders for this scan. Pulling the nozzle further back diminishes the pressure in the differential stage and hence lowers the effusive beam intensity (at the cost of some loss to the available centreline intensity). Note that the nozzle-to-skimmer separations have a zero offset of up to 1 mm due to the difficulty in aligning the nozzle to the fragile skimmer.

123 CHAPTER 4: The Mark II SHeM - Design and Performance

The reductions in the beam path length to optimise the count rate would amount to very little if the effusive beam were once again to overpower the primary beam. In order to determine whether the enhancement to the pump speed for the differential stage was effective, horizontal scans of the nozzle position across the face of the skimmer for a series of nozzle-to-skimmer separations were performed. Similar scans conducted for the Mark I instrument (see Figure 3.4) revealed a sharp peak due to the free-jet beam emerging only once the nozzle was sufficiently far back from the skimmer to reduce the differential stage pressure. The plot in Figure 4.23 demonstrates that even with the nozzle close to the skimmer there remains a sharp central peak, indicating a significant reduction in the effusive beam strength. The ion gauge for the differential stage confirms the pressure drop, with a maximum helium partial pressure of 4.7 x 10-5 mBar for a 6 mm nozzle-to- skimmer separation (as compared to 1.7 x 10-4 mBar at closest approach for the Mark I system). The 6 mm scan in Figure 4.23 still reveals the presence of the effusive beam in the broad shoulders of the central peak, features observed to diminish as the separation is increased. The size of the effusive beam at separations of 12 and 15 mm is virtually identical, due largely to the helium intensity through the skimmer orifice varying as the inverse-square of the nozzle-to-skimmer separation (see plot in Figure 4.24). As such, pulling the nozzle back further than 12 mm results in a reduction in the beam intensity striking the sample surface with no commensurate benefits. For the majority of the sample surfaces imaged with the Mark II SHeM at the time of writing, the nozzle-to-skimmer separation was set to 11 mm to sit in the ‘ideal’ range defined by Campargue and Beijerinck [94, 95] and minimise the strength of the effusive beam. At this separation, a 200 bar beam at ~298K would result in a differential stage helium partial pressure of approximately 2.9 x 10-5 mBar, a marked improvement on the previous instrument (remembering that the differential stage ion gauge in the Mark I underestimates the local pressure near the pinhole by a significant margin).

124 CHAPTER 4: The Mark II SHeM - Design and Performance

Figure 4.24 Plot of maximum differential chamber helium partial pressure as a function of nozzle position. Helium through the skimmer orifice is inversely proportional to the square of the nozzle-to-skimmer separation, as shown by the quality of the inverse-square fit (blue dashed line, R2 = 0.999).

4.6. Conclusions

The Mark II scanning helium microscope design uses the knowledge gained from the planning, construction, and operation of the Mark I instrument in order to build a much more capable system. Key improvements to the count rate, strength of the effusive beam, and sample scanning remedy the main faults with the previous design, and a host of smaller alterations and fine tuning make for an instrument truly capable of benefiting from the advantages of neutral helium microscopy. The following chapter examines a series of materials studies motivated by a desire to further understand the instrument, as well as the contrast offered by the technique.

125 CHAPTER 4: The Mark II SHeM - Design and Performance

126 CHAPTER 5: The Mark II SHeM – Experimental Studies

CHAPTER 5

THE MARK II SHEM – EXPERIMENTAL STUDIES

With the second iteration of the instrument constructed and confirmed to be functioning as intended, focus could shift to understanding the specifics of the technique. This chapter details such investigations, beginning with the imaging of simple samples of a well-known geometry in order to look further into image formation within the SHeM. Such experiments also revealed an unexpected possibility: the potential for stereoscopic imaging of a sample surface. The modular nature of the Mark II instrument allowed for work carrying on from the Mark I concerning the nature and size of the effusive beam contribution. Following this, a study is made of each of the three broad categories of contrast mechanism available to scanning helium microscopy, including the first observations of chemical contrast. Finally, an initial foray into understanding the optics of the microscope is made, in particular a look at the effects of the source parameters and downstream apertures on the resultant image. Data collection for the experimental work shown in this chapter was conducted with the aid of Joel Martens, Kirren Thompson, and Matthew Barr (with some of the work also appearing in Barr’s thesis [32]). All subsequent analysis of the data performed by the author.

Figure 5.1 SHeM micrograph of a hexagonal TEM grid produced with the Mark II scanning helium microscope using a 3 micron step size (scale bar is 100 microns in length). The helium beam strikes the sample from the right side of the image, with the detector aperture sitting to the left. Suspension of the grid off the substrate with carbon tape yields the strong helium shadows observed beneath the grid.

127 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.1 Image Formation

As mentioned in the previous chapter, the move to linear stages incorporating closed loop control for sample movement provided not only the solution to the skew artefacts that hampered the former instrument, but also precise control over the step size used for every scan. A MATLAB script was developed to pull the stage parameters from a micrograph datafile and translate each scan into a Gwyddion (an open source scanning probe microscopy analysis suite) compatible file to allow for distance measurement and other post processing as would be expected for a scanned probe microscopy. All micrographs produced by the Mark II SHeM as shown in this thesis have been generated in Gwyddion in such a manner. In the following micrographs, unless otherwise indicated, the helium beam would strike the sample from the right and the detector aperture would sit to the left.

TEM grids of multiple shapes and sizes were used in the initial testing of the SHeM in order to confirm the expected behaviour, as well as to check distances. Figure 5.2 below shows matched optical and SHeM micrographs of a section of TEM grid (Ted Pella part number 8GC90) mounted to a sample slide with a central hole via carbon tape (black material visible in the bottom of the optical image). From the information provided by the manufacturer, the pitch of the grid should be approximately 282 microns in size. Measurements of the pitch of the grid from the neutral helium micrograph from multiple grid spaces and in multiple directions yielded an average value of (280 ± 25) microns (error taken as twice the step size for the scan), confirming the distance values within the Gwyddion file output from the scan.

Figure 5.2 Matched optical (left) and neutral helium (right) micrographs of a copper TEM grid (Ted Pella part number: 8GC90) mounted to a sample slide with a central hole. SHeM micrograph produced using a step size of 12.5 um; scale bar is 200 microns in length. The pitch of the grid (indicated by doubled headed arrow) as measured from the micrograph is (280 ± 25) microns – well matched to the expected value of 282 microns as provided by the manufacturer.

128 CHAPTER 5: The Mark II SHeM – Experimental Studies

With the microscope confirmed to be reproducing images of samples and reliable distances able to be extracted from the produced micrographs, more definite confirmation of the image formation process was sought. A sample was required not only with well-defined features in a plane (as with the TEM grids), but simple planes at known angles. The solution proved to be silicon nitride membranes as typically used for substrates in x-ray microscopy. Norcada part number NX5025Z consists of a 5 millimetre square silicon frame of 200 micron thickness, in the centre of which sits a 250 micron square silicon nitride membrane, 10 nanometres thick. Figure 5.3 shows the window dimensions, as well as the SHeM micrograph of the same region after the frame was mounted flat to a carbon dot. As expected, the SHeM image displays the same projection distortion due to its scattering geometry as was seen from the Mark I instrument. The top surface of the silicon frame, parallel to the base plane, retains its square shape with all sides measuring 550 microns in length (within experimental error for the scan - once again taken as twice the step size - namely ± 20 micron). The sides of the window sloping down to meet the membrane are a different story; depending on their orientation with regards to the source-detector plane, they can be seen to have different lengths (comparison of the arrows marked ‘1’ and ‘2’ in the figure).

Figure 5.3 SHeM micrograph (10 micron step size, scale bar 200 microns in length) of the central portion of a Norcada silicon nitride x-ray window (part number NX5025Z), with dimensions of the window shown to the right. The projection distortion of the image along the plane of the beam and detector (horizontal plane in the SHeM micrograph) is clear. While the square outer edge of the frame remains unchanged (as a feature parallel to the base plane), the size of the inclined planes down to the actual membrane in the center are different depending on orientation. The inclined planes are steeper than the incident beam, and as such a portion of the square membrane is shadowed and thus blocked from view.

129 CHAPTER 5: The Mark II SHeM – Experimental Studies

Note that the relative brightness of the membrane as compared to the top of the frame is thought to be on account of multiple scattering / pumping effects for the depression, as well as sub-resolution scattering (the silicon nitride membranes are manufactured specifically to be flat and smooth for the purposes of x-ray microscopy).

Based on the geometry of the window, the distance from the outer frame edge to the membrane should appear to be 150 microns. Along the vertical axis (for example, arrow ‘2’), this distance is confirmed in the SHeM micrographs, while in the horizontal axis (arrow ‘1’) there has been a change in length due to the inclination of the plane with respect to the beam. Knowledge of the sample geometry allows us to determine that this face will be 250 microns in length and inclined with respect to the base plane at an angle of 53o (to the nearest degree). By our previously defined formula (equation 3.4), we get a scaling factor of 1.4 and hence an apparent length of 350 microns. The length ‘1’ from Figure 5.3 is (350 ± 20) micron, and so we have excellent agreement with our defined relationship for projection distortion along this direction. It should also be noted that as the angle of the walls down to the membrane is greater than the incident atoms of the beam, part of the surface is obstructed from the beam. The results of the shadowing are readily apparent in the ‘cropped’ appearance of the ordinarily square central membrane.

Figure 5.4 Left: SHeM micrograph of a sugar crystal adhered to a carbon dot (500 micron scale bar, 6 micron step size). Right: CAD render of a replica crystal with light source and camera positioned in the same arrangement as in the SHeM. Note however that the camera and light source had to be swapped in order to produce the shadow as in the SHeM image (shadowing versus masking).

As the final step in the understanding of basic image formation within the SHeM, a fully three dimensional sample was needed, but still one with clearly defined geometry. Sugar crystals belong to the monoclinic crystal system [116], and dependant on the impurities present form clear geometric shapes of a size

130 CHAPTER 5: The Mark II SHeM – Experimental Studies

amenable to producing detailed images with the SHeM. To this end, a raw sugar crystal was adhered to a carbon dot and scanned (Figure 5.4). As a point of comparison, the crystal was recreated in Autodesk Inventor and then the microscope geometry replicated in the rendering suite – the 3D object was placed on a flat plane, a spotlight added at 45 degrees to the plane normal, and the viewpoint similarly locked to 45 degrees on the opposing side of the normal. Comparing the micrograph with the 3D render, we can see good overall agreement between the two in terms of the shape of the crystal and the produced shadow on the substrate. Based on the sides of the crystal visible in the SHeM micrograph, we can conclude that they are steeper relative to the substrate than the incident beam.

The most interesting aspect of the image is the relative intensities from each of the crystal’s faces. Figure 5.5 shows the micrograph again, but with the various sides labelled numerically. We can assume predominantly diffuse scattering from the sugar crystal, with SEM imaging confirming the presence of copious asperities well below the micron scale. As such, the scattered helium will be emitted largely in a cosine distribution centered about the normal to that surface. The top of the crystal (‘1’) will then reflect the most into the detector. It should be noted that count rate comparisons between the top of the crystal and the carbonaceous background

(carbon dot surface) are identical to within RMS noise. \

Figure 5.5 SHeM micrograph of a sugar crystal adhered to a carbon dot (500 um scale bar, 6 micron step size). The relative intensities of the various faces make sense if we consider the helium to scatter predominantly diffusely from the crystal surface. The brightness of side ‘4’ as compared to ‘3’ - despite also being obscured from the detector aperture - is due to multiple scattering events with the carbon substrate.

131 CHAPTER 5: The Mark II SHeM – Experimental Studies

The crystal faces indicated by ‘2’ have smaller mean intensity values as compared to the top face, as dictated by their relative tilts towards the detector aperture. The darkest face in the image (‘3’) has its surface normal facing almost directly back into the incident beam, explaining the minimal helium making its way to the quadrupole detector. Interestingly, the face below this one (‘4’) is brighter. Considering that this face is completely hidden from view of the detector, the only logical explanation is multiple scattering events contributing enough signal to raise the apparent intensity of this face above that of ‘3’. Helium from the incident beam strikes the face, is scattered diffusely back towards the carbon substrate, and a second scattering event leads to some of this helium making its way into the detector volume (demonstrated in Figure 5.6).

Figure 5.6 Assuming predominantly diffuse scattering from the sample surface, the schematic illustration shows how each crystal face (labelled as in Figure 5.5) results in different amounts of helium able to make it to the detector. While face ‘3’ reflects helium back towards the incoming beam, ‘4’ has a much greater opportunity to cause multiple scattering events with the assistance of the substrate, leading to a greater intensity in the final micrograph.

The net result of this investigation into image formation within the SHeM is twofold. Firstly, it helps those viewing the produced micrographs to make intuitive sense of the resultant image, a quality which extends to mounting samples for maximum effectiveness. Similar work was necessary for many other emerging microscopies [117-121], with the SHeM no different. The second outcome is that it offers the potential for deriving something additional from the micrographs - information regarding the surface geometry. Once the intensity from a given surface plane is able to be predicted, it could then be used to extract angles, heights, and other similar data. In other modern microscopies such feats are generally accomplished through forms of three dimensional imaging (a natural evolution of the study of image formation within the instrument).

132 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.1.1 Stereoscopic Imaging

During the initial testing of the new instrument, a series of scans were performed at different working distances – a parameter that was not easily accessible in the previous incarnation of the SHeM. In an experiment to confirm the specular imaging position and examine the effects of working distance on image quality, the sample shown in Figure 5.2 (TEM grid stuck down via carbon tape to a slide with a central hole) was used. A comparison of the produced images appears in Figure 5.7, with the primary difference being a change in the observed count rate. For this particular pair of scans, the longer working distance meant a count rate drop of around 20%, and it should also be noted that no change to signal-to-background was observed (taking the background from the top right of the image, where the helium is allowed to pass through the slide).

Figure 5.7 (a, b) SHeM micrographs of a section of TEM grid (such as shown in the optical micrograph (c)) acquired at working distances of 1.98 and 3.40 millimetres respectively. Note that the range of intensity values has been set identically for both micrographs to allow direct comparison. The masked region caused by the spar in the SHeM images (see arrows) can be observed to shift position, and details such as the shallow depression along the center of each of the TEM grid spars (diamond arrowhead in (a), and visible as a difference in shading in the optical micrograph (c)) become more evident at smaller working distances.

However, there were additional, unexpected changes. In Figure 5.7, we see an enhancement to some topological features at short working distances (in particular, the appearance of the shallow depression along the center of the grid spars), and the masked region caused by the raised grid can be observed to change angle. The cause of these shifts lies in the scattering geometry - by changing the working distance, we are also altering the effective detector angle (see Figure 5.8 below). For the particular pinhole plate in use for these experiments, the specular imaging position would occur at a working distance of 2.86 millimetres. In order to find this

133 CHAPTER 5: The Mark II SHeM – Experimental Studies

position, a blank sample slide was carefully brought up to touch to the pinhole plate, then withdrawn an appropriate distance in Z using the closed loop controllers.

Detector Aperture

Source Aperture

Sample Position

Figure 5.8 Schematic illustrating the effect of different working distances on the scattering geometry. Moving the sample closer to the pinhole plate is achieved by varying the Z-position - by then adjusting the X-stage similarly, we can bring the same feature back into line with the beam. As the different coloured cones suggest, the alteration of working distance changes the relative position of the detector aperture, as well as its acceptance angle.

To fully illustrate the changes occurring, a number of micrographs of the same region of the surface were conducted over a large range of working distances, the results of which are shown in Figure 5.9. At the closest working distances, the edges of the TEM grid spars stand out quite starkly, with areas of lower intensity along the centers indicating a shallow indentation. Looking at the background levels via the transmissive area in the top right, we can see that closest approaches also raise the amount of diffuse helium in the images. The exact cause of this increase is likely to be some combination of the reduced pumping in the vicinity of the pinhole plate (exacerbated by the presence of the effusive beam), and higher levels of multiple scattering events. Total count rates are higher the smaller the working distance, with the exception of Figure 5.9 (a) – at closest approach, the reduction in effective detector angle becomes significant enough to overcome the increase in counts due to the shorter path length. Moving further back towards specular, we see the background signal drop and the shadow tracking to the left of the image. Additionally, the raised edges of the grid are no longer as evident. Pushing past specular, the count rates degrade further, allowing the noise present to start to dominate and cause deterioration in the image quality.

134 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.9 SHeM micrographs of a TEM grid suspended over the edge of a section of carbon tape conducted at different working distances (calculated from known sample to pinhole separations), namely (a) 0.74, (b) 1.44, (c) 2.15, (d) 2.86 (specular), (e) 3.56, (f) 4.27, (g) 4.98, (h) 5.69, and (i) 6.39 millimetres. As the effective detector angle changes, the masked area of the sample surface by the grid spar can be seen to shift, giving a rudimentary form of 3D imaging.

135 CHAPTER 5: The Mark II SHeM – Experimental Studies

Pushing a sample close to the pinhole artificially raises the count rate, but at the expense of similarly raising the background levels present in the produced micrographs (and hence a loss of useful contrast). The inverse is also true – larger working distances means better background rejection but lower count rates, making noise a bigger concern. The specular position can then be seen to be a good balance between the multiple effects, with the ‘cleanest’ image produced. With regards to the appearance of additional topological features at small working distances, we return to a relationship previously defined governing topological contrast (assuming diffuse elastic scattering from a microscopically rough specimen). Diffuse scattering contrast between two planes inclined by ±δ from a mean plane whose normal is itself inclined at an angle of θ with respect to the detector direction will be given by equation 1.4, restated here for clarity:

퐶 = 푡푎푛(휃) 푡푎푛(훿). (Eq. 5.1)

Practical aspects of the scattering geometry then enforce that -90o < θ < 90o and 0o < δ < 90o. Note that we can simplify the former condition down to 0o < θ < 90o, understanding that the difference between positive and negative values of θ will simply be an inversion in topological contrast in the produced image – that is, different sides of features will be directed at/away from the detector. We will also consider an additional condition in order to maximise the information collected from a scan, namely that θ + δ < 90o. If this condition is broken, there is no longer a line of sight to the detector from where the beam strikes the sample surface, resulting in occlusion (masking or shadowing, dependant on the incident beam direction). Figure 5.10 below shows a plot of the topological contrast using this definition for a range of values of θ and δ. Based on the design of the pinhole plate and working distance, with a sample at specular the Mark II SHeM detector aperture would span a range from approximately 33o - 53.5o with respect to the sample normal. This range of θ values has been added to Figure 5.10.

136 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.10 Plot of the magnitude of the topological contrast as given by equation 5.1 for a range of values of θ and δ. Note that when the condition that θ + δ < 90o is broken (i.e.: the detector line-of-sight is blocked - top right half of the plot), the contrast has been set to zero for clarity. The region between the blue dot- dashed lines indicate the range of θ values for the Mark II SHeM with the sample at specular.

It is apparent that large θ values - that is, moving the detector toward a grazing angle, such as is done by reducing the working distance - will increase the observable contrast for a given angular mismatch δ. Additionally, for large θ values the maximum observable angular mismatch between the planes before the onset of occlusion is reduced. Changing the working distance can then be seen to directly affect the size of the observable contrast, thus explaining the appearance of the raised TEM grid spar edges at small working distances. As we pull the sample back, the intensity difference dips below the noise level and the feature thus disappears.

By changing the apparent detector angle, we effectively force the system to provide multiple viewpoints of the sample – hence the change to the area of the sample masked by the spar. The SHeM images in Figure 5.9 then represent the first

137 CHAPTER 5: The Mark II SHeM – Experimental Studies

attempts at three-dimensional imaging using a neutral atom microscope. The use of two (or more) perspectives of the same area to add depth to a produced image is termed a stereoscopic reconstruction [122]. The expansion of such functionality to the SHeM is an intriguing avenue for pulling further information from a sample. In addition to the benefits of the neutral probe, stereoscopic reconstructions potentially allow for the recovery of a height map of the surface under illumination through the use of more advanced computational approaches.

The series of images in Figure 5.9 has set the stage for significant further work in this field. The obvious limitations in changing the working distance means that in order to capitalise on the idea, the same effect must be achieved through sample rotation (thus preserving the count rates, background levels, and noise levels). Recent work by Barr [32] on the Mark II SHeM achieved stereoscopic pairs of images by manually mounting a sample at two set angles relative to the detector. Refining this idea, Myles [123] has now designed a replacement sample mount for the instrument capable of two different modes of 3D imaging. Beyond being a novel way to view the sample surface, precise knowledge of the imaging angles will allow such image sets to provide a height map reconstruction of the sample in question, allowing scanning helium microscopy to extract further detail from scans.

Figure 5.11 (a) Anaglyph of a sugar crystal as built from two SHeM micrographs as imaged using (b) a new sample mount (CAD render) for the Mark II SHeM designed by Myles [123]. The sample mount allows for two modes of 3D imaging and is currently undergoing testing at the time of writing.

138 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.2 Effusive Beam Effects

5.2.1 Mark II SHeM Effusive Beam Contribution

The constricted volume and low pump rate in the differential stage in the Mark I design meant that an undesirably high helium partial pressure formed behind the pinhole. As previously noted, the Mark II design reduces the size of the resultant effusive beam – see Figure 4.23 – but is not be able to remove it entirely. It was then of interest to look at the effect of the secondary beam on the produced images, and if possible, find further ways of reducing its impact on the instrument.

Just as in the Mark I, the partial helium pressure and shape of the pinhole aperture will mean that the effusive beam will take on a shape very similar to that of a cosine distribution. The working distance of the current instrument then leads to the produced effusive beam having a large FWHM (3.6 mm), meaning that the intensity contribution would be close to constant across a typical scanned area. Such a broad background negatively affects image generation in a number of ways. Beyond concerns of resolution (as seen with the Mark I instrument, the sharp supersonic beam must dominate in order to resolve the desired features), the secondary beam will also work to reduce the available contrast. Image contrast is dependent on the beam striking a small area of the sample for a given pixel; contributions from other regions of the sample surface will act to wash out details. The effusive beam strikes a wide area simultaneously, meaning that the detected signal is a combination of a number of interactions between the probe and sample surface. As such, it constitutes an extra helium term without adding to the contrast- producing signal. In the same way, the effusive beam will also degrade the image’s signal-to-background and signal-to-noise ratios.

One of the first checks performed in order to investigate image quality was the signal-to-background ratio. For the Mark I system, shadowed regions of a sample surface were used to determine a representative S/B ratio of 0.15 (representative due to the value being dependant on the sample choice and stagnation conditions of the detector). Similar experiments were performed with the initial configuration of the Mark II instrument using holes in the sample slide (for example, Figure 5.2 and Figure 5.18). With no surface to reflect helium into the detector, these holes yield excellent positions to take a measurement of the background levels for the instrument. Across multiple samples, the S/B ratio averaged somewhere in the range 0.36-0.41, a marked improvement to image quality over the Mark I.

139 CHAPTER 5: The Mark II SHeM – Experimental Studies

As a more intuitive comparison of the image quality with the Mark I, a section of a curved TEM grid was imaged with and without a strong secondary effusive contribution to the beam profile. In order to do so, first an image was collected using a nozzle-to-skimmer separation of 11 mm (maximising the dominance of the supersonic free-jet beam). With a 200 bar beam and the aforementioned nozzle- to-skimmer separation, the differential chamber saw a helium partial pressure of 3.2 × 10−5 millibars. Next, the intensity of the secondary beam was raised by backfilling the differential stage with an additional 4.0 × 10−4 millibars of helium gas.

Figure 5.12 (a) SHeM micrograph of a TEM grid adhered to a cleaned silicon wafer with a small piece of carbon tape (20 micron step size used; scale bar represents 1 millimetre). (b) and (c) show sections of the grid – indicated by the square shown in (a) – imaged using a beam with differing ratios of supersonic to effusive beam. (b) shows the grid as imaged with the raw supersonic beam and effusive contribution native to the SHeM with a nozzle-to-skimmer separation of 11 mm, while (c) includes a more significant effusive beam contribution. To cause the latter, the differential chamber had an additional 4.0 x 10-4 millibars of diffuse helium added. Note that (b) and (c) use the same range of intensities – centred on the median intensity for each micrograph to account for the differences in raw count rate – to allow for direct comparison. The increase in the secondary beam causes an increase in noise present and reduces the available contrast.

140 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.12 (a) shows the large area scan of the TEM grid, while (b) and (c) show the micrographs taken before and after the enhancement of the secondary beam. The presence of the secondary beam can been seen in the raw counts observed in each image - Figure 5.12 (b) has less than half the counts per pixel as compared to (c). Note that in order to provide a means for a direct comparison of image quality given this difference, (b) and (c) have been rendered over the same range of intensities, but centred on the median intensity for each micrograph.

The comparison of the same region with and without the artificial enhancement to the secondary beam makes the detrimental effects quite clear – there is obviously a greater amount of noise, and the edges of the spars are not as distinct. As discussed, the effusive beam forms a strong broad background without contributing to the useful contrast. While we cannot determine a true S/B ratio for the sample, taking the minimum and maximum count rates for each images produces a reasonable estimate. The calculated values are compared in Table 5.1, with a satisfactory match to those determined previously for the Mark I and Mark II instruments respectively. We can also calculate the Michelson contrast for the carbon tape towards the left hand side of each image (used so as to avoid the region to the right, occluded by the raised TEM grid). Taking the contrast as the difference between the bright silicon substrate and the dark edges of the grid, we see that image (b) has close to three times the Michelson contrast of (c); a substantial improvement. A comparison of the noise levels requires a little more thought due to the greater helium count rate for the image with the enhanced effusive contribution. Signal-to-noise ratios will tend to favour higher count rate images, but it is clear that the higher intensities in Figure 5.12 (c) are not valuable signal. Thus for this calculation, the minimum count rate in each image was set to zero (acting as a background subtraction to leave only the useful contrast), and a signal-to-noise ratio calculated for the bright areas of silicon. As expected simply by looking at the comparison, the effusive beam more than halves the value, resulting in substantial loss to image quality.

141 CHAPTER 5: The Mark II SHeM – Experimental Studies

Supersonic Beam Supersonic with

Only Effusive Contribution

S/B Estimate 0.41 0. 11

Michelson Contrast 0.148 ± 0.008 0.056 ± 0.004

Signal-to-noise 30.62 ± 8.00 12.12 ± 2.88

Table 5.1 Comparison of the quality of the micrographs presented in Figure 5.12 (b) and (c) due to the influence of the stronger effusive beam. The signal-to- background ratios and Michelson contrasts are produced by comparing the brightest and darkest features within the image, namely the silicon substrate and the TEM grid edges. It should also be noted that for the signal-to-noise ratio, the minimum intensity in both micrographs was set to zero (as discussed in the text).

5.2.2 Further Improvements

The size of the background helium partial pressure behind the pinhole is determined in part by the geometry of the pinhole plate. With the obvious improvements to image quality possible through control over this region, a slight modification of the modular pinhole plate was attempted. Previous illustrations of the pinhole plate have shown a perfect conical bore for simplicity. The actual pinhole plates were achieved through a series of step drills of reducing size to approximate the desired cross-section as shown in Figure 5.13 (b). The resultant shape thus presents surface area perpendicular to the incident beam (and hence will cause undesirable backscattering), and a much greater restriction to the available pumping. Material from the sides of the pathway through the pinhole plate was removed in order to broaden the beam entry for reasons of pumping, and to remove as much of the exposed faces as possible. Figure 5.13 (c) shows the change in shape made to the pinhole plate, illustrating the considerable difference. Conductance calculations (see calculations for a tapered pipe of circular cross- section in Roth [124]) indicate the pumping available to this volume improves by 10-20 times than that available to the original design, meaning a significant reduction to the size of the effusive beam present. The modified pinhole plate had a 5 micron pinhole glued into position in exactly the same manner as the previous plate.

142 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.13 (a) Photograph of the beam entry into the pinhole plate in its original form. (b) and (c) show schematic diagrams of the beam entry to the pinhole plate before and after the alteration to improve pumping around the pinhole (respectively). The bored-out pinhole plate improves the pumping around the pinhole and results in a reduced effusive beam accompanying the main supersonic free-jet beam.

Much as with the TEM grid image in Figure 5.12 above, a direct comparison of the change in image quality due to the alterations to the pinhole plate was desired. To this end, a section of a salt crystal was imaged first with the original pinhole plate, then with the one with expanded pumping using identical beam and detector conditions (Figure 5.14). The count rates for the modified pinhole plate are lower than the original by approximately 25%, but with an improvement to the image quality. Contrast ratios are higher for the altered plate, especially apparent when looking in the hollow at the center of the crystal. Michelson contrast ratios for the carbon tape as compared to the background observed within the hollow yield values of 0.16 and 0.28 for Figure 5.14 (a) and (b) respectively. Occlusion of the detector aperture from the incident beam means that a good estimate of the signal- to-background ratio can be made here: (0.39 ± 0.09) for the original plate, and (0.79 ± 0.11) for the modified version. The comparison supports the conclusion that the effusive beam acts to reduce the effective contrast by contributing signal from neighbouring sample areas. Positions on the surface that, if imaged with a more directed beam spot, would appear black or white instead also receive signal from less stark features nearby and thus become more grey.

143 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.14 SHeM micrographs of a salt crystal illustrating improvement to the pumping around the pinhole. (a) uses the original pinhole plate, and was collected using an 8 micron step size, while the pinhole plate in (b) had been opened up to improve pumping (7 micron step size). Both images were collected at otherwise identical beam and detector conditions, with colour-bar indicating the relative count rates. Scale bars are 250 microns in length.

In an effort to further illustrate the effects of the effusive beam on image contrast, histograms of the pixel intensities for each of the micrographs in Figure 5.14 were generated (Figure 5.15). The plots clearly show the change in intensities due to the improvement to the pumping around the pinhole. Comparing the histograms and the original images, we see that there are peaks in the count rate distributions that match certain features – for instance, the peak towards the lower intensities is that of the occluded features (hollow in the middle of the crystal, masked region of carbon tape, etc.), while the peak at high intensities corresponds to the carbon tape substrate which has a high reflectivity. Both features are examples of what would be a single intensity in an idealised sample, but are broadened due to the realities of a true sample geometry. However, it is noticeable that the histogram peaks are sharper for the modified pinhole plate, especially when the background is subtracted to give a more direct comparison of the actual contrast (Figure 5.15 (b)). The effusive beam ‘smears’ out the contrast from these features, reducing the effective contrast.

144 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.15 (a) Histogram of the pixel intensities for the SHeM micrographs collected of the salt crystal with the original (red) and modified (blue) pinhole plates (as seen in Figure 5.14). The difference in count rates is immediately apparent. (b) is the same data but after each set of intensities have been normalised to zero background, allowing a more direct comparison of the shape of the intensity distributions and revealing the broadening of peaks associated to certain sample features by the effusive beam.

To determine the overall improvements made on the instrument with respect to the secondary beam, a section of honey bee (Apis Melifera) eye was imaged with a range of possible effusive beam contributions. The three micrographs in Figure 5.16 illustrate the sizable effect on image quality of changes to the pumping around the pinhole plate. While all three images have identical beam and detector settings, the following conditions differ:

 (a) Original Mark II SHeM pinhole plate with a 5 micron aperture, but the bypass between source and differential chambers open to allow an extra 4.0 x 10-4 millibars of diffuse helium to populate the latter. The result emulates the Mark I instrument (a much stronger effusive beam).

 (b) Original Mark II SHeM pinhole plate, bypass closed.

 (c) Modified Mark II SHeM pinhole plate, bypass closed. Improved pumping around the pinhole leads to a reduced effusive beam.

145 CHAPTER 5: The Mark II SHeM – Experimental Studies

While the improvement is evident through inspection of the micrographs in Figure 5.16, there are several particular areas of note. The improvement to contrast through the more selective nature of the beam means that details in the bottom left corner of the micrograph - namely part of the bee head next to the compound eye - become discernible in (c) relative to (a). Similar effects can be seen on the pollen caught on one of the hairs to the top right of the image – as the secondary effusive beam contribution is lowered, the shape can be clearly resolved. The curvature of the eye is also more apparent, due to the reduction in background and again a more selective beam (meaning that the mean plane changes which give rise topological contrast are more localised). Table 5.2 summarises the differences between the three micrographs in terms of the signal-to-background ratio and Michelson contrast available.

(a) (b) (c)

S/B 0.13 0.38 0.56

Michelson Contrast 0.06 0.16 0.22

Table 5.2 Image quality metrics pulled from the micrographs shown in Figure 5.16. Due to the organic nature of the sample, finding larger areas with sufficient pixels to pull viable statistics from proved impossible, hence the lack of errors for these measurements.

146 CHAPTER 5: The Mark II SHeM – Experimental Studies

The degradation of image quality is of particular importance when attempting to investigate the more novel mechanisms, such as chemical contrast. As HAS studies have shown, the amount of helium scattered inelastically is 2-3 orders of magnitude lower than that scattered elastically [24], meaning that the contrast mechanisms dependant on such an interaction will be much weaker than topological contrast. As such, minimising the effusive beam contribution is critical if chemical contrast is to be observed.

147 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.3 Contrast Mechanisms

5.3.1 Topological

Numerous SHeM Mark II micrographs of samples have been shown thus far in this chapter, all of which exhibit strong topological contrast. It is worth reiterating that no special preparation was performed for any of these samples (or those to follow) – in the majority of instances, the specimen of interest was simply adhered to a sample slide via carbon dot and then placed straight into the sample chamber. The neutral character of the helium probe particle means that metallic coatings such as those required to prevent surface charging in SEM and TEM are unnecessary. By contrast, imaging without such coatings in electron microscopy leads to deleterious charging effects, as can be seen in Figure 1.1

With the previous discussion of image formation previously already covering aspects of the influence of topology (shadowing, masking, etc.), this section will instead focus on the ability of neutral helium to avoid the potential for damage, even with fragile materials. The delicate nature of the probe particle, and lack of conductive coating means that samples which would sustain damage under the energetic beams can be routinely imaged. For example, many of the organic films necessary to fabricate organic electronics are easily destroyed by SEM/TEM/X-ray imaging, especially given that typical film thicknesses range from tens to a few hundred nanometers. An opportunity to image such a film emerged with samples being prepared for a trip to the Advanced Light Source in Berkeley where they would undergo scanning transmission x-ray microscopy or STXM. Incidentally, STXM is a technique where great care must be taken to avoid both obvious damage such as outright destruction of the film by the X-ray beam, as well more insidious damage including damage to selected intermolecular bonds.

A C60 film was spun down onto glass to a thickness of approximately 80 nanometres (as determined by profilometry), then floated off and pieces caught on a TEM grid. As the holes that form in the sections of film are of interest, it was decided to try to maximise the available contrast and highlight their location. To do so, the suspension of the sample above the sample slide - used in images such as Figure 5.1 - can be taken further and an aspect of transmission microscopy added. The first spatially resolved images produced with neutral helium were through a transmission microscopy [58]; however, the surface sensitivity of the helium atoms means that in pure transmission the recorded intensities will either be zero or maximal (neglecting secondary effects such as background gas levels, multiple

148 CHAPTER 5: The Mark II SHeM – Experimental Studies

scatterings, detector noise, and diffraction). By suspending the sample mounted in the SHeM over a gap in the sample slide (see Figure 5.17) we are able to take advantage of the large contrast available through letting the helium passing through the gaps, while still collecting reflection information.

Figure 5.17 Sample mounting geometry for enhancing contrast in thin films (side and top views). The sample – in this illustration a TEM grid - is suspended over a gap in the mounting plate, allowing helium that passes through the sample to quickly become part of the background.

The produced SHeM micrographs can be seen in Figure 5.18; as desired, the gaps in the sections of the film show up quite clearly against the background. The image still retains useful information from the reflected helium, namely the position of the pieces of film relative to the TEM grid they are supported on. For instance, it can be seen that the film is sitting on the top side of the TEM grid, and positions where the film has curled outward. No changes to the organic film, or any other sample, were observed no matter how long they spent exposed to the helium beam.

Figure 5.18 SHeM micrographs of a section of TEM grid onto which portions of a spin-cast C60 organic film have been floated. (a) Survey scan (200 micron scale bar, 15 micron steps) of the grid showing how it has been placed over a hole in the sample slide to allow helium to pass through to form a transmission-like image. Boxed region indicates the scan area shown in (b), a higher resolution image (50 micron scale bar, 2 micron step size) showcasing the increased contrast afforded. While the

149 CHAPTER 5: The Mark II SHeM – Experimental Studies

transmission-like mounting immediately reveals all the holes that form in the spin- cast film, the reflection mode image contains information that would be either very difficult or impossible to determine from a purely transmissive image (for example, the placement of each piece of film on the grid surface, especially for the curled sections in the top right of the image).

Figure 5.19 below shows sections of a wing from the butterfly species Tirumala Hamata as imaged by the Mark II SHeM, including a matched optical microscope image (Leica M205C) to serve as a point of comparison. Butterfly wings are made up of two chitinous membranes sandwiched together with a network of veins in- between [125, 126]. Covering the outside of each wing are thousands of tiny, delicate scales, also formed from chitin. In order to image the scales via electron microscopy they must be first coated with a conductive film, else suffer charging effects.

150 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.19 Matched micrographs of a butterfly wing (Tirumala hamata). (a) Reflection optical micrograph (Leica M205 C), scale bar 600 micron. (b) Scanning helium micrograph as imaged with the Mark II SHeM with 8 micron steps, scale bar 600 micron. (c) SHeM micrograph of region indicated by square in (b) using 4 micron steps, scale bar 100 micron. (d) SHeM micrograph of another area of the wing taken with 4 micron steps, highlighting the shape of the wing scale (scale bar 50 micron).

151 CHAPTER 5: The Mark II SHeM – Experimental Studies

The optical micrograph in Figure 5.19 (a) shows the difficulty in keeping the entire wing section in focus over such a large image area. Conversely, the SHeM micrograph displays excellent depth of field despite estimations that the various ridges and folds of the wing result in a sample height varying over almost 2 millimetres. The mean plane differences across the wing surface cause the changes in helium intensity, with shadowing and/or masking provided by the scales making for an intuitive image. The ripples in the wing membrane are quite evident despite the covering of scales, and the veins running through the top half of the wing are clearly identified, with the changing angles of the scales jutting out from the surface picked up through occlusions. Comparison between the optical and SHeM micrographs also reveals that the colour difference in scales for this particular species is also a morphological difference, with the blue scales being flatter and more streamlined as compared to the black [126]. In Figure 5.19 (c) a position where a scale has previously been removed from the wing can be observed, leaving behind a curved divot in the membrane.

As a further demonstration of imaging delicate biological samples, optical and SHeM micrographs were taken of a section of a wing from the honey bee species Apis mellifera. The complex folds of the membrane of the wing are the almost indiscernible in the optical image due to the transparency of the membrane material and the range of sample plane heights (in Figure 5.20 (a), the distance from the sample slide to wing top is ~1.5 mm – see Section 5.4 for more detail on the instrument depth-of-field). However, all of these features are readily apparent in the SHeM image, especially the way in which the membrane is stretched taught over the support structure underneath and the relative heights of the different features. The absolute surface sensitivity of the helium atoms means that only features on the top side of the wing are observable. For example, masking of the incident helium beam is visible from both the translucent hairs on the wing surface in Figure 5.20 (d), and where the wing rests on the substrate. Thus, SHeM produces intuitive images of biological samples with no sample preparation required and no risk of beam damage to the substrate.

152 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.20 Comparison of reflection optical (Leica M205 C) (a,c) and SHeM (b,d) micrographs of a honey bee wing (Apis mellifera) as an example of topological contrast. Bottom images taken from the square region indicated in (a). SHeM micrographs collected with (b) 8.75 micron and (d) 2 micron steps. Scale bars are 500 and 50 microns for (a, b) and (c, d) respectively.

153 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.3.2 Chemical

Topological contrast will be the dominant mechanism dictating the observed variations in helium signal from a sample surface. However, the composition and local atomic character of a sample surface should also give rise to differences which might be exploited in the form of image contrast. For example, the highly- corrugated isopotentials of a semiconductor should result in more diffuse scattering than would the delocalised electrons of a metal [31]. Image contrast will also arise due to inelastic scattering of helium due to energy exchange with surface vibrations (phonons), known to be highly dependent upon sample composition. While such interactions will be present for any noble gas species, an atomic beam of helium is an ideal probe of such processes since its small size and mass minimises the lattice displacement and hence will excite or de-excite the largest number of vibrational modes possible [30, 34]. Based on the decades of work with helium atom scattering, this contrast mechanism had been postulated in literature for SHeM [31], but hitherto unobserved.

The nature of chemical contrast, namely the strength of the effect relative to topological effects, means that the initial sample choice was critical. A sample with a variety of different surface planes will exhibit competing mechanisms, and care must be taken to avoid coatings that would leave the surface all the same material (a problem with the tin spheres attempted on the Mark I system). To achieve an unambiguous observation of any chemical contrast effects, it was decided to manufacture a flat plane with two dissimilar materials in an easily identifiable pattern, namely that of the University of Newcastle logo. With the aid of the ACT node of the Australian National Fabrication Facilities (ANFF) network, electron beam lithography was used on a silicon (100) substrate with native oxide layer to deposit first a 3 nanometre thick titanium wetting layer, followed by different metallic films. 5 samples in total were prepared: 40 nanometre thick logos made of gold, nickel, platinum and chromium, and an additional 15 nanometre thick gold logo.

By keeping the metallic films so thin, it was hoped that there would be almost zero topological effects, especially given the techniques seeming indifference to height variations, provided the planes are parallel. Choosing a metal-semiconductor interface gave the maximum probability for observation of a significant (observable) difference in the reflectivity, and the metals of choice are all noble metals or form thin passivating layers, reducing the chances for complex comparisons to the established predictions in the literature.

154 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.21 (a) SEM micrograph of a 40 nanometre thick gold University of Newcastle logo produced via electron beam lithography at the ACT node of the ANFF network. Identical logos in other metals (chromium, nickel and platinum) were produced for the purposes of an investigation into chemical contrast with scanning helium microscopy. (b) SHeM micrograph of the same sample. Image was collected with a 12 second dwell per pixel, and a 7 micron step size. Despite the clear appearance of the logo, the contrast is much smaller than any of the scans shown previously. Scale bars in both micrographs are 50 microns in length.

The 40 nanometre thick gold sample was the first of the logos to be imaged, and indeed the characteristic shape showed up bright against the darker silicon background (Figure 5.21 (b) shows the results of one of the early locator scans). It should be noted that the expectations regarding the relative strength of the contrast mechanism proved to be true. In some spots on the silicon chips, small areas of photoresist remained after the lithography resulting in occluded areas of the produced SHeM micrographs which rendered the logos invisible by comparison. Comparison of the pixel intensities in the micrograph shown in Figure 5.21 (b) reveal a count rate difference of approximately 90 per second between the gold and the silicon, and calculations of the Michelson contrast yield a value around 0.006. The contrast is appreciably lower than any observed for topological features in either the Mark I and Mark II instruments. The reduced contrast meant that collection times for the logo samples had to be increased to upwards of ten seconds per pixel to ensure acceptable image quality, leading to scans lasting over the course of days. In some instances, horizontal banding could be observed in the produced micrographs – small variations in the detector sensitivity over the long scan times led to count rate changes that were discernible against the flat background. Careful masking of the silicon in Gwyddion allowed these effects to be minimised and the micrographs to be freed of the artefacts as much as possible.

155 CHAPTER 5: The Mark II SHeM – Experimental Studies

With the gold logo proving to be visible in SHeM images, attention now turned to confirming that the origin of the contrast was chemical in nature. The difference in the size of the contrast was a good first indicator. As a more direct comparison, Figure 5.22 (a) shows a micrograph of the 15 nanometre thick gold logo, on which a dust particle had landed during sample preparation. The fortuitous occurrence allowed an image to be produced that directly showcases the difference in intensities, consistent with the metal-semiconductor contrast arising due to a less common scattering mechanism.

Figure 5.22 SHeM micrographs of the 15 nanometre thick gold University of Newcastle logo partially obscured by a piece of dust. (a) Full image taken with the sample at specular, while (b) shows a section of the sample at the same position. (c,d) Small region of the sample at 500 and 1,000 μm further back from the pinhole respectively. Scale bar 50 microns.

The unique sample also allowed a further confirmation to be made. The scattering geometry of the SHeM is designed to maximise the potential to observe contrast deriving from inelastic effects, as interactions with phonons will tend to eject scattered atoms out of the specular channel. As discussed previously, shifting the sample position allows the effective detector position to be altered, and hence the sample to be imaged out of specular. Figure 5.22 (b) through (d) show a progression as the gold logo was moved back from the specular position in 500 micron increments. Although the detector acceptance angle in the current apparatus is large, the distance moved is sufficient to place the sample specular position outside the detector’s acceptance cone. While the contrast between the gold and silicon diminishes with increasing sample distance from the specular position, there is no trend in the contrast between the dust particle and underlying

156 CHAPTER 5: The Mark II SHeM – Experimental Studies

silicon as a function of sample position. This observation indicates that while the contrast in the dust particle is purely topological in nature, the change in contrast for the gold logo cannot be due to a simple step height or mean plane topological feature.

It is possible that a consistent difference in surface roughness between the gold and silicon (sub-resolution topological features) could explain the observed contrast. To eliminate this possibility, atomic force microscope (AFM) studies were conducted using a Cypher scanning probe microscope with sub-nanometre resolution in both lateral and axial directions to determine the degree of surface irregularity in both materials. It was found that the gold layer had a root mean square (RMS) roughness of 2.8 nm, while the silicon yielded a value of 1.2 nm. A coarser surface would, in general, be expected to cause a higher degree of diffuse scattering away from specular, and thus would appear darker in a SHeM micrograph; the inverse to the observed results in Figure 5.21 and Figure 5.22. As such, it was deemed unlikely that the difference in roughness is the cause of the observed contrast.

With the gold logos producing the expected contrast and behaviour, it was time to confirm a difference in relative intensities with the other metal/semiconductor combinations. All four 40 nanometre thick logos were mounted to a single sample slide, with extreme care being taken to ensure the silicon substrates were as consistently flat as possible. In the produced SHeM micrographs, not only was there distinct contrast between the metals and the silicon substrate, but also between each of the different metallic species. Figure 5.23 shows a composite image of the four different logos, with the incident helium intensities normalised using the mean silicon signal as the reference.

As before for the 15 nm thick gold logo, AFM was employed in order to rule out topological features below the resolution of the instrument being the cause of the observed contrast. For the gold, nickel, platinum, chromium and silicon, the measured mean RMS roughnesses were 2.1, 0.9, 1.6, 2.6 and 1.4 nm, respectively. Given that the observed contrast does not follow the roughness trend in the AFM data and that the same wetting layer was used for each sample, it seems unlikely that changes in diffuse scattering arising from nanometre scale surface irregularities can explain the contrast differences observed across the four samples.

157 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.23 SHeM micrographs of the 40 nanometre thick University logos in different metals on pieces of the same silicon substrate. Clockwise from top left: (a) gold (b) nickel (c) platinum and (d) chromium. Scale bar is 50 microns in length. In all four images, the intensities have been normalised relative to the silicon background in order to make a direct comparison between the metals possible. Additionally, some of the collected micrographs have been rotated, thus changing the directions so far associated with the incident beam and detector pathway.

Having thus confirmed the spread of backscattered helium (and hence the signal collected by the detector aperture) varied across the different materials, it was decided to compare the results against the limited theoretical grounding on the subject. As discussed in Chapter 1, the Debye–Waller factor is used to describe the variation in specular (‘in-plane’) reflectivity (I/I0) via interactions with phonons. In an attempt to produce a mathematical expression for potential chemical contrast, MacLaren et. al. [19] employed the Michelson contrast definition (eq. 1.2) and the DWF formulation shown in equation 1.5, restated here for clarity:

2 24 푚 푇 (퐸푖 푐표푠 휙푖 + 퐷) 퐼 − 2 = 푒 푀 푘 훩퐷 . (Eq. 5.2) 퐼0

158 CHAPTER 5: The Mark II SHeM – Experimental Studies

Equation 5.2 describes an atomic beam of energy Ei, mass m, incident on a surface

of atomic mass M at angle ϕi, temperature T and Debye temperature ΘD. D is the potential well depth of the interaction of the helium atom with the surface and k is the Boltzmann constant. The expected contrast then between two ideal, smooth,

crystalline materials due to lattice vibrations is based on their atomic masses (MA

and MB) and surface Debye temperatures (ΘA and ΘB):

훼 1 1 퐶 = 푡푎푛ℎ { ( − )} (Eq. 5.3) 2 푀퐴훩퐴 푀퐵훩퐵 2 24 푚 푇 (퐸푖 푐표푠 휙푖 + 퐷) 훼 = (Eq. 5.4) 푘 For the Mark II SHeM, we can determine the α parameter using T = 294K,

o Ei = 64 meV (originating from a 298K beam at 200 bar), ϕi = 45 , m = 4 AMU, and assuming a well depth of 6 meV for all materials [19]. We can consequently determine a theoretical prediction of the contrast for each of the experimental systems, as shown below in Table 5.3. While the contrast for nickel and platinum at least qualitatively matches that of the model, the gold and chromium is inverted as compared to the prediction. The result is perhaps not unexpected; early work with the scattering of helium from surfaces (which attempted to verify the DWF for gas–surface interactions) produced consistent agreement with the DWF for some systems, yet for others it was wholly inadequate [45]. Attempts have been made in recent times to redefine the DWF in terms of quantum gas–surface interactions [46], but it is clear that the complex vibrational behaviour of materials remains an active area of research within the field of surface scattering. It is also important to note that these observations do not rely on the specific form of the metal surface layer. The metals utilized in this study were chosen to reduce the complexity of any subsequent comparisons to theory. Nevertheless, oxide or other physisorbed layers of contamination will certainly be present in our ex-situ prepared samples. However, it is clear that chemical contrast from the underlying material is still evident in the micrographs. In addition, the presence of adsorbates on the sample surface does not necessarily prevent such interactions since the surface charge density is impacted by the motion of atoms buried deep below the surface and the incident helium atoms can probe these subsurface resonances (see Section 1.3). In principle, therefore, the helium–electron–phonon coupling can provide chemical contrast with which to characterize a surface even in the presence of multiple adsorbate layers.

159 CHAPTER 5: The Mark II SHeM – Experimental Studies

Sample System Experimental Contrast Modelled Contrast

Gold on Silicon 0.015 ± 0.005 -0.48

Nickel on Silicon 0.005 ± 0.003 0.05

Platinum on Silicon 0.001 ± 0.003 0.02

Chromium on Silicon -0.009 ± 0.004 0.26

Table 5.3 Comparison of the experimentally determined Michelson contrast for each of the metallic logos on silicon and the predicted contrast using the model of MacLaren et. al. [19].

While the predictions based on the DWF do not match the experimental results for the metallic films, they do suggest a further experiment in order to help confirm the origin of the contrast. Varying the energy of the incident helium beam (through control over the temperature of the free-jet expansions stagnation volume) should alter the available contrast, provided it arises due to inelastic effects. The 15 nanometre gold logo sample was scanned with a beam energies in the range of 83 – 21 meV (387 – 97K stagnation temperatures), the results of which are shown below in the first row of Figure 5.24. As a control study, a sample with strong topological contrast - our traditional TEM grid on a silicon wafer substrate - was also imaged using identical conditions across the same range of mean beam energies. For the TEM grid, there was no observed trend in the Michelson contrast as a function of temperature; however, for the logo the contrast decays as the beam energy moves to lower and lower values, as predicted by theory for inelastic scattering (see equation 5.2).

However, the count rate of the instrument is directly tied to the beam temperature (changes to the centreline intensity, and possible alterations to the alignment to the downstream optics as the beam apparatus shifts with temperature), with lower energy beams yielding a significantly lower count rate at the detector. As such, the trend in the contrast with temperature could simply be a result of the changes in the signal-to-noise ratio as a function of beam intensity. In order to rule this out as the underlying cause, a simple simulation was produced using the MATLAB script that was employed to determine acceptable count rate levels for the Mark I instrument (see Chapter 2). For a given greyscale image, the script used a designated average count rate and contrast level between the two materials to produce a simulated SHeM image, with a controllable degree of noise added in

160 CHAPTER 5: The Mark II SHeM – Experimental Studies

(nominally set to Poisson shot noise as determined by the average count rate). An empirical baseline level of contrast was determined from the 66 meV SHeM image (produced from a stagnation volume at close to room temperature). Then, using the average count rate and noise levels from each SHeM micrograph (the latter pulled from known silicon regions), simulated versions of each of the experimental images were produced, as shown in Figure 5.24 (simulation 1).

Figure 5.24 SHeM micrographs demonstrating the dependence of the observed contrast for the 15 nanometre thick gold-on-silicon sample with helium mean beam energy. All scans were performed with the sample in the specular imaging position, at room temperature (294K), and using a 200 bar stagnation pressure beam. Utilising the heating/cooling system for the beam described previously, beam energies of (a) 83, (b) 72, (c) 66, (d) 42, and (e) 21 meV were achieved. Below the experimental results are simulated micrographs, testing whether count rate changes were the cause of the observed differences. Simulation 1 involved matching the average count rate and noise levels for each original micrograph, with the contrast level constant as determined from the 66 meV image. Simulation 2 went a step further, and included a variable contrast between the gold and silicon. Comparisons of the experimental and simulated micrographs led to the conclusion that there was an additional contrast mechanism at play.

The results of the simulation showed that the location of the gold film did indeed become more difficult to observe as the mean beam energy (and associated count rate) were lowered. However, the change was not great enough to stop the logo being easily identifiable – a marked difference to the experimental results. A second simulation was then conducted whereby not only were the changes in mean count rate and noise matched to the experimental results, but also the

161 CHAPTER 5: The Mark II SHeM – Experimental Studies

contrast between materials. Figure 5.24 (simulation 2) shows the simulation is well matched to the experimental data set, achieved only through a direct reduction in material contrast.

Contrast in the SHeM is a result of the angular distribution of the backscattered helium, with local vibrational modes influencing the degree of inelastic scattering, while surface topography dictates the diffuse elastic scattering. As suggested by the DWF, the inelastic scattering is strongly dependant on beam energy (in comparison to the weak influence for elastic scattering through the much smaller associated change in momentum). Given the observed trend in the micrographs as a function of stagnation temperature could not be explained by a loss of signal, but instead a significant loss of contrast, it was concluded that the dominant mechanism present was a difference in inelastic scattering between the gold and silicon.

The discussed experimental observations constitute the first observations of contrast for a neutral atomic beam microscope due to the local composition of the sample surface. While very promising in terms of the prospects for future imaging opportunities, there remains a sizable amount of work to understand and harness the effect fully. As evidenced by the inability to predict the observed contrast using the DWF as a measure of the degree of inelastic scattering, more work remains in order to understand these helium–electron–phonon interactions. The ability of bulk effects to propagate and influence the electron population at the surface of a material in a sufficient way as to be probed by atom scattering provides tantalising glimpses into SHeM being able to move past even chemical differences, and start to characterise these deeper phonon interactions. Additionally, recent work on the Mark II instrument has shown that much as with HAS, SHeM is sensitive to even very thin layers of metal across a substrate. A silicon dioxide surface known to have height variations of the order of hundreds of nanometres across a lateral distance of several millimetres was masked with a TEM grid and a sub-nanometre layer of gold evaporated using thermal evaporation (6.8 Ångstroms as recorded by the QCM). After removing the mask, the sample was imaged with the SHeM (Figure 5.25) and despite the thickness of the gold layer there was clear contrast between the areas of the surface with and without the metallic film. Attempts thus far to characterise directly the thickness of the film via AFM have failed, in part due to the extreme roughness of the underlying substrate. While still early, such results indicate that the SHeM may be able to exploit the contrast mechanism to examine films otherwise invisible to the majority of other imaging techniques.

162 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.25 SHeM micrograph of a rough silicon oxide surface which had been masked with a TEM grid and then a sub-nanometre thick gold film thermally evaporated. Even with the topographic feature provided by the carbon tape (bottom left corner), the chemical contrast allows for the location of the gold film to be seen easily (darker regions on the flat sample area). Inset shows a section of the masked surface if the count rates are limited to those due to the gold film on the silicon oxide surface. Scale bar 250 microns in length. Micrograph collected in collaboration with Dr. Sabrina Eder.

5.3.3 Diffractive

Given the discussion thus far in this chapter concerning topological and chemical contrasts, it is worth a brief mention of the third contrast mechanism identified in literature: diffractive effects. Under normal scattering conditions, thermal helium atoms have a de Broglie wavelength comparable to typical crystallographic dimensions. As a result, the elastically scattered helium atoms produce diffraction patterns characteristic of the surface corrugation potential, and this angular variation in the scattered helium atom intensity can be exploited as contrast. Additionally, interference effects can occur from coherent scattering from neighbouring atomic layers, a quality that has led to HAS becoming one of the most sensitive techniques for studying thin film growth [24, 25, 30]. Such a contrast mechanism would be an immense boon to the technique, as it potentially provides intensity differences between regions of a sample with identical chemical composition, but different degrees of crystallinity or similar changes to the local order.

While a sample surface may diffract the incident helium into a channel other than the specular, it may not necessarily direct it outside the acceptance angle of the detector, functionally rendering the effect void. To resolve these diffraction

163 CHAPTER 5: The Mark II SHeM – Experimental Studies

patterns, typical HAS apparatus have an angular resolution of less than half a degree [25, 48]. As shown in Figure 5.10, the current iteration of the SHeM has a larger detector aperture to account for the deficiency in terms of helium signal, and as a result the angular distribution of any diffraction peaks will fall within the subtended detector angle therefore precluding the observation of individual diffraction or interference peaks. Consequently, diffractive effects do not presently play a role in the generated contrast. Furthermore, it is well known from prior HAS studies [25] that even small amounts of disordered adsorbates lead to a reduction in the intensity of the helium diffraction peaks, and so sample cleaning would be a necessary requirement. To adapt the SHeM to achieve HAS-like angular resolution, a 3–4 order of magnitude increase in detector sensitivity is required (based on the reduction in detector aperture size), an increase which may be feasible through new technology being researched in this area. A discussion of the requirements for diffractive contrast to play a role in SHeM imaging is included in the following chapter.

5.4 Instrument Optics

With the accelerated development timeframe for the Cambridge SHeM, it was not possible to perform any experiments concerning the nature of the helium spot projected onto the sample surface beyond the simple line scan across an edge as shown in Figure 3.20. While a purely geometric calculation of the spot size projected onto the specular plane is possible, it makes many assumptions - including ignoring the beam profile. Understanding the shape of the beam profile is crucial in establishing the fundamental resolution of the instrument, but also of interest (at least in terms of the instrument design) is the full diameter of the spot. To a first approximation, the resolution is dictated by this spot size, which in turn will be set by the divergence of the helium beam through the final collimating aperture. However, the spot size may be broadened or even shaped by aberrations arising from the finite size of the helium source, aperture diffraction or even the effects of background gas.

The full point spread function (PSF) of the instrument (the response of an imaging system to a point source [89, 127]) can be problematic to find in microscopies where the sample is rastered under a fixed beam, and so instead its cross-section - termed the line spread function (LSF) – is often used [89, 128, 129]. Finding the LSF requires scanning the beam spot across a suitably sharp step edge

164 CHAPTER 5: The Mark II SHeM – Experimental Studies

(henceforth referred to as the ‘knife-edge’) to generate the edge spread function (ESF, also known as the step response of the instrument). Numerical differentiation of the ESF will then yield the LSF.

As has been demonstrated previously [130-132], provided the edge the scan is taken over is sufficiently sharp and well characterised (by SEM/TEM for instance), it is relatively easy to determine an estimate of the line spread function of the instrument and hence the resolution. In many of the higher energy microscopies, finding a knife-edge can be difficult due to the penetration of the probe into and even through some materials [132]. However, the surface sensitive nature of the neutral helium probe simplifies the problem to finding a clean and straight edge sharper than the beam profile (nominally 5 um based on the pinhole). To this end, a silicon (100) wafer with a thick oxide layer was cleaved until a suitably straight section was found.

The subject of beam profiles as produced by skimmed supersonic free-jet expansions is one of much examination, as the final shape is a result of the complex interplay between a large number of variables [34, 50, 55, 56, 94, 95, 133- 135]. The shape is most commonly Gaussian in nature, but the broadness of the distribution can vary significantly. With the pinhole only allowing a small section of the profile to strike the sample surface, the profile could potentially be a small section at the peak of the Gaussian, yielding a fairly constant intensity. To attempt to classify the experimental results, one can imagine two potential ideal profile distributions between which the true profile will sit: a uniform intensity across the entire beam spot (a ‘top hat distribution’) or a typical Gaussian distribution. Figure 5.26 below shows the ideal beam spot for both the top hat and Gaussian distributions, along with the associated beam profile and LSF each would produce.

Scans across the silicon knife-edge in the horizontal axis were performed, with Figure 5.27 showing the resultant ESF along with the corresponding LSF for the SHeM. For this particular scan, the conditions were set to those most commonly used for imaging: a 200 bar beam with a room temperature stagnation volume (~298K), a nozzle-to-skimmer separation of 11 mm, a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns, and a 5 micron pinhole set into an unmodified pinhole plate. It is clear from the shape of the LSF that the beam profile of the instrument has much more in common with a Gaussian than a top hat. Attempting to fit the LSF with a Gaussian function yields good agreement (an R- squared value of 0.984), and from the σ value of the fit we derive a FWHM of (6.9 ± 0.2) microns.

165 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.26 Comparison of the two most likely candidates for the beam profile, namely a top hat distribution (a) and a Gaussian distribution (b). Determining the intensity profile of the beam directly is difficult – instead, by scanning the beam across a feature (typically a sharp edge) in one dimension, the line spread function (LSF) can be found.

Figure 5.27 Edge Spread Function (ESF) and Line Spread Function (LSF) for the Newcastle SHeM for the most common imaging parameters. The ESF is found by scanning the beam spot across a silicon knife edge placed at the specular position (in half micron steps), while the LSF is found by numerically differentiating the ESF. Note that the ESF has had a 3 pixel moving average smoothing filter applied for reasons of noise and the ends padded to simplify the resultant LSF.

166 CHAPTER 5: The Mark II SHeM – Experimental Studies

The shape of the beam profile means that defining the resolution of the instrument simply through the size of the spot produced on the sample surface is perhaps not as valuable as originally intended. As the sample moves underneath the beam, the imaged area will see a much smoother variation in helium intensity, and so a more logical definition of resolution would be to appeal to the literature concerning the separation of two point objects. As the results of knife-edge scans yield an approximately Gaussian profile, we can employ the Rayleigh Criterion as a measure of resolution [136, 137]. The Criterion states that the limit for two point sources to be considered resolved occurs at the point where the principle diffraction maximum of one aligns with the first minimum of the other – for Gaussian LSFs, this corresponds to a dip in the central combined intensity profile of approximately 26.5% [136]. Taking the beam profile shown in Figure 5.27, specifically the Gaussian fit to the LSF, we can convolve it with very basic 2D features and look at the produced profile. For the purposes of defining resolution, the features chosen were two boxcars separated by a gap, the size of which was varied until the desired drop in intensity was observed. As a clear example of the process, Figure 5.28 (a) uses boxcars separated by a gap of 4.8 microns to achieve the Rayleigh Criterion.

The drop in intensity between the point sources corresponds to a Michelson contrast of approximately 0.15. As evidenced by the chemical contrast images shown earlier, such a difference is well below the capabilities of the instrument. Correspondingly, for a better estimate of true capabilities of the Mark II SHeM we will instead employ an alternative two-point resolution criteria, namely the Dawes Limit [137, 138]. Empirically developed by astronomers attempting to resolve double stars (and subsequently applied to microscopy), the Dawes Limit corresponds to a central intensity dip of 3 – 5% [137, 138]. For our purposes, we will classify a 3% drop in the central profile as the requisite condition. As a point of comparison to earlier experimental results, the observed Michelson contrast value

for the evaporated nickel metal film on silicon substrate (CM = 0.005) would map to an intensity loss of 1% (thus making the Dawes Limit appropriate for our purposes). Figure 5.28 (b) once again uses the beam profile derived from the knife-edge scan in Figure 5.27, but this time seeks a 3% intensity drop. A separation of approximately 2.8 microns is required in the simulation, and hence this is taken as the resolution of the instrument.

167 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.28 Result of convolving the Gaussian fit for the LSF from Figure 5.27 with a periodic grid. The grid spacing sets the size of the dip in the central intensity for the convolution, corresponding to different resolution definitions. (a) Rayleigh Criterion (~26.5% difference in intensity), achieved by a separation of 4.8 microns. (b) A grid spacing of 2.8 microns results in the convolution obeying the Dawes Limit (~ 3% difference in intensity). The latter criteria agrees well with the observed contrast for the instrument, and so 2.8 microns is taken as the resolution for the most common imaging parameters used.

It should be noted that the knife-edge scan in question was performed in the horizontal axis, and will hence always be larger than that in the vertical axis due to the scattering geometry. Vertical knife-edge scans showed a similar shape to those in the horizontal axis, and for the same instrument settings the FWHM derived was (5.4 ± 0.2) microns, corresponding to a resolution of approximately 1.5 microns via the Dawes Limit (Rayleigh Criterion: 3.1 microns). Knowledge of the Gaussian profile of the beam also allows for a quantification of the available depth-of-field using the Rayleigh range [139]. From considerations of the pinhole diameter,

168 CHAPTER 5: The Mark II SHeM – Experimental Studies

working distance, and utilising the FWHM for the beam waist, a value of approximately 4.3 millimetres for the Rayleigh range can be derived. The large depth of field compares well with SEM, and helps to explain the intuitive, three- dimensional nature of the produced micrographs.

While no true knife-edge scans were conducted for the Mark I instrument, as shown in Figures 4.19 and 4.20 we may extract some vertical line profiles through specific images to derive an approximation. Taking the result of these linescans as the ESF and performing the same analysis as above (Figure 5.29), we obtain a FWHM of (5.4 ± 0.2) microns and a resolution according to the Dawes Limit of approximately 1.5 microns (Rayleigh Criterion: 3.0 microns).

Figure 5.29 Top: ESF and LSF for the Mark I SHeM, obtained from the vertical profiles shown in Figure 4.19 . The ESF has had a 3 pixel moving average filter applied and the ends padded. The Gaussian function fit to the LSF is in good agreement (R-squared value of 0.991), and yields a FWHM of (5.4 ± 0.2) microns. Bottom: Results of convolving the Gaussian beam profile for the Mark I SHeM with a periodic feature with separation of 1.5 microns to satisfy the Dawes Limit.

169 CHAPTER 5: The Mark II SHeM – Experimental Studies

Both instruments utilise 5 micron pinholes, and while improvements were made to the size of the effusive beam contribution in the Mark II, it is expected to be so broad as to minimally affect the resolution (as noted previously). Such a measurement was a good confirmation of continuity across the different instrument iterations, especially considering the reduction in beamline length involved shifting the relative separation of the optical elements.

Further investigations of the resolution for the Mark II instrument were performed by changing the pinhole that formed the final optical element. As discussed for the effusive beam contribution, two different pinhole plates with 5 micron pinholes had already been fabricated (the original, as well as a modified plate where material had been removed from the beam inlet to improve pumping). Adding to this, a 2 micron pinhole was mounted in a pinhole plate which also had material removed from the inlet channel. For a 200 bar beam with a room temperature stagnation volume (~298K), a nozzle-to-skimmer separation of 11 mm, a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns, scans in the horizontal axis were performed across the same section of the silicon knife-edge, the results of which are shown in Figure 5.30. Note that the data for the original 5 micron pinhole is that same as shown in Figure 5.27.

The change from the original to modified pinhole plate, both using 5 micron pinholes in the silicon nitride membranes, resulted in a very small improvement to the resolution (FWHM values of (6.9 ± 0.2) microns and (6.5 ± 0.3) microns respectively); one which when considering the errors in the fitting could simply be due to noise in the scans. However, the 2 micron pinhole yields a very interesting result, with a FWHM value of (6.1 ± 0.2) microns. While smaller again than the previous two alternatives, the corresponding improvement in resolution is much smaller than might be expected for the relative change in diameter. It should be noted that SEM images of the silicon nitride membrane verified the diameter was as expected. Furthermore, the pinhole was confirmed to be behaving geometrically in terms of the observed count rates – when the modified 5 micron pinhole was swapped out for the 2 micron, the count rates at the detector for an identical beam and sample were reduced by the same ratio as the relative pinhole areas (a factor of 6.25).

170 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.30 Top: Gaussian fits to the LSFs derived from horizontal knife-edge scans for three different pinhole arrangements for the Mark II SHeM. Note that the scans have been aligned to the peak of the Gaussians for ease of comparison. Bottom: Full width half maximum values for the three different pinholes extracted from the Gaussian fits.

To assess the differences more directly in terms of actual scans, SHeM images of the edge of a butterfly wing were obtained using both the 5 micron and 2 micron pinholes – see Figure 5.31. It is obvious when directly comparing the two that the 2 micron pinhole does indeed produce sharper images, with finer details such as the ragged edges of the wings appearing. However, the improvement is not commensurate with the change in pinhole diameter, meaning that there is a limiting factor to the system optics that must be found in order to improve resolution appropriately. The constraint is further emphasised by the loss of counts when moving to the smaller pinhole, evident in the quality of the SHeM micrograph shown

171 CHAPTER 5: The Mark II SHeM – Experimental Studies

in Figure 5.31 (b). All told, the improvement to the sharpness is simply not worth the rise in image noise or extended scan times.

The performance of the two micron pinhole, relative to what might naively be expected, was a strong indication that the SHeM optics are a much more complicated system than originally thought. As a first step towards gaining a better understanding of its complexity, experiments were performed in an attempt to derive some information concerning the virtual source of the instrument – that is, the spatial distribution of the beam which then dictates the effectiveness of the downstream apertures. The spatial extent of the source will thus affect the final dimensions of the helium spot on the sample surface (geometric demagnification of the virtual source).

Figure 5.31 SHeM micrographs of the edge of a butterfly wing as imaged using different pinhole diameters, namely (a) 5 microns and (b) 2 microns. Both scans were conducted with identical beam conditions, and used 1.5 micron steps (scale bars 50 microns in length). While it can be seen that the micrograph utilising the 2 micron pinhole is indeed sharper than the 5 micron, the improvement to resolution was not comparable to the reduction in pinhole diameter. Also note the effects of the reduced intensity in (b), especially the prevalence of noise.

The topic is already one of interest for the field of supersonic free-jet beam sources, especially in relation to more recent work on neutral atom optics designed to focus said beams. The basis for the most popular approach to a formulation of the virtual source is what is known as the ‘sudden freeze’ model of a free-jet expansion. When the atomic or molecular species of choice expands through the nozzle orifice into a rarefied environment, the random thermal energy of the gas is converted through collisions into the translational kinetic energy of the beam. Moving past the nozzle, the collisional frequency between atoms decreases as the rapid expansion results in a drop in the local density, leading to the gas atoms following straight line

172 CHAPTER 5: The Mark II SHeM – Experimental Studies

trajectories typically referred to as ‘streamlines’; any further impacts can be considered small perturbations relative to the primary expansion axis. At the point where the number of collisions has become negligible, the sudden freeze model creates a boundary construct (the ‘quitting surface’) between the region of continuum flow and that of molecular flow. While in reality the transition occurs over a much larger region of space, the boundary does allow the calculation of a wide variety of parameters useful to the study of such beams [34]. In the process of expanding, the velocity distribution of the gas atoms narrows until the reduction in collisional frequency stops the distribution from evolving further. At this point, the temperature of the gas is said to have been ‘frozen in’, which in conjunction with the assumption of the quitting surface gives the sudden freeze model its name.

A definition of the virtual source can be obtained from aspects of the expansion between the nozzle and the quitting surface. Beijerinck and Verster [101] in particular describe a ‘virtual source point’ a few microns in front of the nozzle where all streamlines followed by atoms in the expansion appear to emanate from (see Figure 5.32). We may then define a ‘virtual source plane’ normal to the expansion axis through the virtual source point. Mapping the distribution of all trajectories of atoms past the quitting surface back onto this plane forms the virtual source of atoms in the expansion – our optical source for the instrument. Figure 5.33 shows a schematic representation of this mapping.

173 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.33 Schematic diagram illustrating the concept of the virtual source for a free-jet expansion. By mapping the distribution of trajectories at the quitting surface back to the virtual source plane, we may define the virtual source. The perpendicular velocity distribution of the atoms in the expansion as they leave the quitting surface is the critical parameter describing the resultant size of the virtual source. Image courtesy of Barr [32].

At the quitting surface, the velocities of the atoms along the different axes, typically classified as parallel and perpendicular to the beam expansion axis, have diverged significantly. Accordingly, the sudden freeze model assumes a total velocity distribution approximated by an anisotropic distribution, namely the product of two Maxwellian functions parallel and perpendicular to the streamlines of the expansion. The given definition of the virtual source then becomes dependant on the two velocity distributions; in particular, the virtual source size is a function of the perpendicular beam temperature [140] (and thus the lateral coherence of the beam). It is well established in the literature that the virtual source of a free-jet expansion can be described by a Gaussian distribution [34] – likely the origin of the shape to the observed beam profiles for both iterations of the SHeM discussed thus far. However, the specifics are still an active subject of research, with more sophisticated modelling being undertaken to match theoretical calculations with experimental results. For example, it is known that as collisions decrease during the expansion, the different beam temperatures become increasingly decoupled. The perpendicular axis in particular exhibits non-equilibrium behaviour [140], resulting in the perpendicular speed distribution no longer able to be described by a single Gaussian. While the complexities of describing these inter-atomic dynamics is beyond the scope of this thesis, for the purposes of the following

174 CHAPTER 5: The Mark II SHeM – Experimental Studies

discussion the author wishes to note that the virtual source size has been found to be strongly dependant on the stagnation conditions, including pressure and temperature [34, 50, 51, 94, 95, 140]. With regard to the latter, chilling the beam will act to reduce the perpendicular velocity and thus reduce the size of the virtual source. Also of note is that backscattering into the centreline of the beam is thought to affect the broadness in a meaningful way, and will be especially prevalent at higher stagnation pressures. With regards to what might be expected as a ‘typical’ virtual source size, published experimental work for another instrument employing a free-jet beam source utilising a 10 micron nozzle reported a width of (180 ± 9) microns with stagnation conditions of 171 bar and 320 K [141].

As a quick first investigation of the virtual source for the instrument, horizontal knife-edge scans were collected as a function of the nozzle-to-skimmer separation

ZNS. Conditions for the experiments included a 200 bar, room temperature stagnation volume beam source, sampled by a Type 2 Beam Dynamics skimmer with nominal diameter of 120 microns, and the original 5 micron diameter pinhole. The FWHMs of the Gaussian fits to the ESFs can be seen in Figure 5.34 with the

resolution of the instrument appearing to have no dependence on ZNS.

Figure 5.34 FWHM values extracted from the Gaussian fitting for horizontal knife- edge scans conducted at various nozzle-to-skimmer separations. All scans were conducted using a 200 bar beam with room temperature stagnation volume and a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns. No trend was observed in the data, indicating the independence of instrument resolution from the nozzle-to-skimmer separation.

175 CHAPTER 5: The Mark II SHeM – Experimental Studies

The seeming lack of a change in instrument resolution as a function of nozzle position along the beam axis goes against what has been previously reported in the literature. Using geometric optics, Witham et. al. [86] defines the optical source for such a microscope as the quitting surface of the free-jet expansion. As a means to investigate the validity of such a hypothesis, it was decided to vary the temperature of the stagnation volume and hence alter the velocity distribution of the beam. Additionally, this change to the velocity distribution will also shift the quitting surface location – the colder the stagnation, the greater the distance before inter-atomic collisions tend towards zero. Miller [34] gives that the quitting surface of a supersonic free-jet expansion in the sudden freeze model can be approximated by a hemispherical cap of diameter Dqs:

1 푀 퐷 = 2 푑 ( ∞ )훾−1 (Eq. 5.5) 푞푠 3.232 where 푀∞ is the terminal Mach number, γ the ratio of specific heats (helium: 5/3), and d the nozzle diameter. The Mach number can be derived from the speed ratio S of the expansion:

훾 푆 = √ 푀 , (Eq. 5.6) 2 with the terminal speed ratio estimated from:

4 0.495 4 1 3 푆∞ = 0.778 [ 푃0 푑 (9.57푥10 ( ) )] . (Eq. 5.7) 푇0

For a 200 bar beam and nozzle-to-skimmer separation of 9 mm, three different beam stagnation volume temperatures were achieved using the counterflow cooling system incorporated into the Mark II beam source; vertical knife-edge scans were conducted at each. Table 5.4 shows the FWHM values as derived from Gaussian fits to the LSFs for each knife-edge scan set, along with the calculated quitting surface diameter and terminal speed ratio for each temperature. Figure 5.35 plots the FWHM values, along with a linear fit to the experimental data.

176 CHAPTER 5: The Mark II SHeM – Experimental Studies

Calculated Stagnation Experimental Calculated Terminal Quitting Surface Temperature (K) FWHM (um) Speed Ratio Diameter (um)

97 ± 3 4.7 ± 0.2 7460 153

298 ± 1 5.4 ± 0.2 2450 73

380 ± 1 5.7 ± 0.3 1930 62

Table 5.4 FWHM values derived from Gaussian fits to vertical knife-edge scans for different beam temperatures. As a point of comparison, the approximate size of the quitting surface for each temperature (given a 200 bar stagnation pressure, and 10 micron nozzle diameter) as given by Miller [34] is included, as well as the terminal speed ratio.

Figure 5.35 FWHM values extracted from the Gaussian fitting for vertical knife- edge scans conducted at different beam stagnation temperatures. All scans were conducted using a 200 bar beam and a Beam Dynamics Type 2 skimmer with nominal diameter of 120 microns. The width of the beam profile increases with increasing stagnation temperature, as would be expected considering the monochromaticity of the beam source and hence any potential broadening effects. Data is fit well (R2 = 0.999) by a linear function as indicated by the dotted line.

177 CHAPTER 5: The Mark II SHeM – Experimental Studies

By geometric optics (see below), if the size of the virtual source were to increase then so should the final spot size on the sample surface of the instrument. As can be seen for the calculated values, chilling the stagnation volume significantly increases the size of the quitting surface, along with the monochromaticity of the beam (as determined by the speed ratio). The experimental data however demonstrates an improvement to resolution for colder temperatures, indicating that the quitting surface does not form the virtual source for the SHeM. The absence of any change with nozzle-to-skimmer separation, along with the performance of the 2 micron pinhole, would then seem to indicate that the size of the skimmer is limiting the instrument performance. Should the skimmer always sample only a portion of the virtual source, then its size would become the critical dimension – essentially making the skimmer the optical source for the SHeM.

In order to determine whether the skimmer size was controlling our final instrument resolution, a study was devised whereby three different skimmer orifices would be employed and further knife-edge scans conducted. The skimmers in question were Beam Dynamics Type 1 skimmers with orifice diameters of 100, 200 and 510 microns. Note that the Type 1 skimmers are smaller in overall profile than the Type 2 model (such as was used in the experiments described up to this point). In particular, the height of the skimmer from base to tip is some 6 mm shorter for the Type 1 model, meaning that the relative placement of the optical elements is shifted. As such, the 120 micron Type 2 skimmer was not included in the data series (further discussion on this point will follow). With a 200 bar beam stagnation at room temperature, a nozzle-to-skimmer separation of 15 mm and the modified 5 micron pinhole installed, horizontal knife-edge scans were conducted for each skimmer, the results of which are shown in Figure 5.36 and Table 5.5.

178 CHAPTER 5: The Mark II SHeM – Experimental Studies

Skimmer Orifice Experimental Gaussian Fit Resolution Diameter (microns) FWHM (um) R2 value Estimate (microns)

100 5.67 ± 0.4 0.980 1.7

200 7.7 ± 0.2 0.991 3.5

510 19.8 ± 0.6 0.977 15.1

Table 5.5 FWHM values derived from Gaussian fits to horizontal knife-edge scans for different skimmer orifices. For the experiments in question, system employed a 200 bar beam, room temperature stagnation volume, a nozzle-to-skimmer separation of 15 mm. and the modified 5 micron pinhole plate. Resolution estimate based on the Dawes limit as previous detailed.

Figure 5.36 FWHM values extracted from the Gaussian fitting for horizontal knife- edge scans performed with a range of skimmer sizes. Beam Dynamics Type 1 skimmers with nominal diameters of 100, 200 and 510 microns were used, with a nozzle-to-skimmer separation of 15 millimetres. Beam stagnation was 200 bar, room temperature, and the modified 5 micron pinhole plate formed the final optical element. A direct dependence of the beam profile width with respect to skimmer diameter can be seen, indicating the skimmer forms the restricting element in terms of resolution.

179 CHAPTER 5: The Mark II SHeM – Experimental Studies

Figure 5.37 Experimental LSF data and associated Gaussian fits for horizontal knife-edge scans performed with Beam Dynamics Type 1 skimmers with nominal diameters of (a) 100, (b) 200, and (c) 510 microns.

180 CHAPTER 5: The Mark II SHeM – Experimental Studies

The results of the experiment show that the larger skimmers give rise to much larger beam profiles and hence reduced resolutions. Taken in context of the previous results, the author concludes that in the current instrument the skimmer is acting as the limiting element optically, and in order to push resolutions below those so far observed smaller skimmers must be implemented. Unfortunately, skimmers in the form factor thus far employed in the Mark II SHeM (ie: Type 1 and Type 2 models from Beam Dynamics) in sizes below 100 microns are not commercially available, making the change more difficult. Traditionally, skimmers in this range were manufactured in house, often by drawing glass pipettes and then polishing the tips to form what are known as ‘microskimmers’ [23, 56, 133]. Fabrication of appropriate microskimmers is a difficult process, and due to the restricitve orifice size and often extended height of the produced skimmer much greater care must be taken with alignment to ensure the centerline of the beam is allowed to pass through undisturbed.

The minimum aperture diameter for a microskimmer is sub-micron, and is limited only by the skill of the person manufacturing the piece. Smaller skimmers will reduce the intensity observed further downstream, and considering the manufacture process it would be prudent to have some estimate of the required size (as opposed to optimising purely experimentally). As a first pass, we turn to a geometric model of the system optics first developed by Witham et. al. [86]. Simple geometry and knowledge of the various apertures along the beamline (see Figure 5.38) allow an expression for the maximum spot size (S) on the sample surface to be produced should the atoms follow straight-line trajectories, namely:

푊퐷 푊퐷+퐿 푆 = 퐷 ( ) + 퐷 ( ). (Eq. 5.8) 푆 퐿 푃 퐿

Ds is the virtual source diameter, DP the pinhole diameter, L the source to aperture separation, and WD the working distance for the instrument. While originally used to estimate sample spot size based on the quitting surface of the beam being the virtual source, knowing that we are currently most critically limited by the skimmer diameter we can adapt the model to suit the SHeM. Based on the use of the Type

1 Beam Dynamics skimmers, we take L = 43 mm, WD = 2.86 mm, DP = 5 um, and

vary Ds to calculate the expected spot size. Note that with the shape to our beam, the maximum spot size is not a good match to the FWHM as has been derived from prior knife-edge scans. To this end, the full width tenth maximum (FWTM) has instead been calculated in much the same way as the FWHM from the Gaussian

181 CHAPTER 5: The Mark II SHeM – Experimental Studies

fits to the derivative of the knife-edge scans. A comparison of the experimental and model results is shown below in Figure 5.39.

Figure 5.38 Schematic of the beam geometry used for equation 5.8 describing the size of the spot produced on the sample surface by geometric optics. For the Mark II SHeM, the virtual source diameter is taken as the skimmer orifice diameter, in contrast to the original use of the formula whereby the quitting surface was used. Figure after Witham [86].

Figure 5.39 Plot of the spot diameter for the Mark II SHeM as found experimentally with knife-edge scans and via a geometric optics model for different skimmer diameters. The experimental spot size was taken as the full width tenth maximum as calculated from the Gaussian fitting to the LSFs as described previously. Both data series have been fitted with linear trends as a point of comparison.

182 CHAPTER 5: The Mark II SHeM – Experimental Studies

A comparison of the geometric model with the experimental results initially looks promising: the model values are slightly larger than the FWTM values, the trend is very similar (linear fits to both series yield slopes within 3%), and adjusting the relevant parameters to account for the 120 micron diameter Type 2 skimmer also produces decent agreement. However, this is likely more serendipitous than an accurate reflection of the underlying physics. If configured for the 120 micron Type 2 skimmer, swapping out the 5 micron pinhole for the 2 micron version in the model produces a spot size of 9.1 microns – much smaller than the FWTM of (11.1 ± 0.4) microns as found from the fit to the knife-edge scans. Furthermore, the earlier experiments whereby the beam stagnation temperature was varied show that the resolution can change with no alteration of the physical dimensions. While perhaps not useful quantitatively, the model (along with the results in Figure 5.36) shows that reductions in skimmer aperture can have a large effect on the produced spot size.

With the inherent complexity of the SHeM optics issue, it is likely that a much more encompassing set of experiments needs to be performed – a characterisation of the resolution for multiple skimmers, pinholes, beam stagnation pressures and temperatures, and potentially even optical element separations. Given the hefty time commitment this is likely to entail, modelling the pathway of helium atoms through the system more comprehensively would also be a suitable alternative, one which would be of much use when designing the next iteration of the instrument.

Some immediate options do present themselves. It is obvious that in order to push the resolution forward, the next iteration of the instrument must transition to microskimmers, a move that would hopefully allow the smaller pinhole sizes to make the desired impact. Additional options to improve the resolution would include chilling the beam stagnation volume for routine scans or extending the beam path length in order to help with the monochromaticity. It is worth noting that these options have been observed to drop the count rates at the detector, and thus we arrive again at the trade-off between signal and resolution that has accompanied spatially resolved helium atom scattering from the very beginning. It is likely that improvements to neutral atom detection, a subject covered in the next chapter, are needed to truly advance in a meaningful fashion.

183 CHAPTER 5: The Mark II SHeM – Experimental Studies

5.5 Conclusions

The performance improvements of the Mark II SHeM in comparison to the Mark I instrument have allowed it to progress appreciably our understanding of the technique. The hypotheses regarding image formation developed using the Mark I were able to be confirmed and expanded upon; enabling further information concerning sample geometry to be extracted from the SHeM scans. The potential for 3D imaging in particular remains an alluring prospect. In addition to this refinement of our approach to topological contrast, the Mark II instrument demonstrated the first evidence of chemical contrast. Deriving contrast from the local chemical environment for a sample, despite the presence of an overlayer of physisorbed species such as oxygen and water, opens the door to helium microscopy adding spectroscopic analyses to its available toolkit.

There still remains much work to be done, not just for the Mark II specifically, but the field as a whole. The modular nature of the system means that the Mark II will be able to readily adapt to new technological advances, unlocking further avenues of research in the process. A discussion of the nature of these technological advancements and how they might be incorporated into the existing designs is the subject of the next chapter.

184 CHAPTER 6: The Future for Scanning Helium Microscopy

CHAPTER 6

THE FUTURE FOR SCANNING HELIUM MICROSCOPY

The suggestion to utilise the unique properties of neutral helium as a probe particle in order to produce a novel microscopy was set down in literature as early as 1970. From an article in Science titled ‘Molecular Microscopy: Fundamental Limitations’ by J.R. Breedlove Jr and G.T. Trammell [35]:

“On the basis of estimates of molecular damage caused by the observation process, it is concluded that molecular microscopy of biological molecules in which the individual atoms are resolved is impossible with an electron or x-ray microscope. Microscopes that use low-energy helium atoms or neutrons as illuminants may be capable as serving as ultimate biomolecular microscopes.”

Interest in such ‘atom microscopes’ peaked again in the 1990’s with the development of new and more reliable atom optics [59, 142-148], and in the span of 20 years that idea has now matured into a global field of research [19, 31, 56, 58, 86, 87, 149-154]. As of the time of writing, there are four separate systems [58, 86, 149, 150] which have produced what can be termed SHeM micrographs (two of which are detailed earlier in this thesis). However, with the bulk of the work in the area thus far focused on overcoming technical limitations, the proliferation of working instruments leads to an obvious question: where do we go from here? In order to continue the forward momentum the emerging field has gathered, such an important subject must be given careful thought. The considerations for the next generation of scanning helium microscopes extend beyond simple refinement of the designs detailed thus far in this thesis. Much as was done in moving from the Cambridge prototype to the Newcastle instrument; one could focus on improving known issues – modifications to the beamline length, detector performance, differential pumping stage, and other similar enhancements would indeed result in improved scan performance. However, in light of our increased understanding of the available contrast mechanisms and their interaction with the limitations of the

185 CHAPTER 6: The Future for Scanning Helium Microscopy

instrument, approaching the question from a more fundamental position is a valuable exercise.

With multiple instruments and three different scattering geometries capable of producing images, it would be sensible then to investigate the relative merits of each design. Following on from the current state of the field is then a discussion of the specific practical requirements associated with each of the contrast mechanisms. Focusing the discussion in this manner allows future designs to better exploit the available signal through judicious design changes, or even the potential incorporation of technology currently under development, and thus provide some sense of the challenges and opportunities that lie ahead.

6.1 Alternate Scattering Geometries

6.1.1 Transmission SHeM

In 2007, Koch et. al. [58] produced the very first micrographs utilising neutral helium by employing a zone plate to focus down a beam onto a sample and then collecting a transmission image. Figure 6.1 below shows images collected of a hexagonal TEM grid with scan times not dissimilar to those of the Mark I SHeM (total dwell times of between 8 and 14 seconds per pixel, with a total scan time of approximately 14 hours for each image).

Figure 6.1 The first neutral helium images as produced by Koch et. al. [58] of a hexagonal TEM grid. Micrograph (a) was produced with a beam focused down to a 3 micron spot and 8 seconds collection per pixel, while the zoomed region (b) used a 2 micron spot and 14 seconds collection time per pixel.

The detection stage sits directly behind the sample (relative to the incident beam – see Figure 6.2) meaning that the images are formed in a manner analogous to transmission optical or electron microscopy. However, unlike other transmission techniques where some portion of the beam may be transmitted directly through the sample, the special properties of the helium probe particle mean the produced micrographs are much more ‘binary’ in nature. The extreme surface sensitivity

186 CHAPTER 6: The Future for Scanning Helium Microscopy

makes the result somewhat less useful than the analogues – either the helium beam may pass directly through a hole in the sample and onto the detector (bright regions in the produced micrographs), or it will strike the surface and hence be occluded. As a result, transmission imaging with neutral helium is a technique suited to porous thin films which might sustain damage under energetic probes – for example, cellulose acetate membranes, photonic crystals, or sheets of graphene - but the majority of other samples would likely be better served with reflection imaging.

Figure 6.2 Schematic of the optical layout for the first neutral helium images [58]. By having the helium beam pass directly through the sample to strike the detector an image may be produced in a manner analogous to transmission mode optical. The resolution of the produced TEM grid image was 1.9 ± 0.1 microns, limited by the chromatic aberrations of the zone plate acting as the focusing element.

It should be noted that the optical arrangement shown in Figure 6.2 was not intended as a final arrangement for imaging; rather, it served as a proof of concept for the use of neutral helium for the purposes of microscopy. The demonstration of the power of Fresnel zone plates as a means to focus the atoms down to a workable spot size also provided critical information regarding their implementation. As an adapted atom scatterer, the length of the beamline is large – in particular, the zone plate is set 1492 mm from the nozzle, and the sample a further 680-840 mm from the zone plate (variable to account for focus position). Beyond allowing ample opportunity for multiple differential pumping stages and hence reduction of the background helium levels (especially important considering

187 CHAPTER 6: The Future for Scanning Helium Microscopy

the transmission percentages of traditional zone plates), the distance from the source now factors into the performance of the optical element. The zone plate in question consisted of 2700 zones with an outer zone width of 50 nanometres, and in practice produced a spot with a FWHM of 1.9 ± 0.1 microns using a microskimmer of 1.2 micron diameter with a 190 bar beam stagnation pressure.

The expected diameter of the focused spot in this configuration was (0.62 ± 0.05) microns, indicating the presence of significant broadening. Zone plates in such applications will (in general) suffer from chromatic aberrations, namely longitudinal chromatic aberrations resulting from the finite velocity spread of the supersonic free-jet expansion and transverse chromatic aberrations from off-axis atoms striking the zone plate. Increasing the source stagnation pressure has the effect of raising the terminal speed ratio and hence monochromaticity of the beam [58, 153], but control of the size and shape of the virtual source will dictate the extent to which the transverse chromatic aberrations affect the resultant spot size. Extending the distance from source to optical element reduces the number of off-axis helium atoms being diffracted into the focused spot, thus improving the performance of the zone plate at the cost of total intensity.

It is the conventional question of intensity that forms the main obstruction for the move to reflection mode imaging with a zone plate. For the images in Figure 6.1 (a) the signal was quoted as approximately 80 counts/second on a background of 110 counts/second, showing the influence of the other focus order spots on the signal-to-background present when imaging. Further work by the research group has led to significant refinement of the zone plate designs, including the addition of a post zone plate order sorting aperture capable of removing a significant portion of the remaining zero order beam contribution to the background. To try to maximise the intensity from their instrument, the group has produced theoretical work concerning what they have termed Neutral Atom Microscopy (NEMI for short), including a paper where they model the performance of a neutral helium microscope employing a zone plate as the optical element [155]. The results of the simulations show that improving the performance of the instrument (in terms of intensity for a set focal spot of 0.9 micron FWHM) requires bringing the nozzle to zone plate distance from 1528 millimetres down to 555 millimeters, and reducing the radius of the zone plate. In the author’s own words: “To summarise: the theoretical best helium microscope design is a compact microscope with a relatively large zone plate, combining the closeness to the atom source with a large angle of collection” [155]. Such work would seem to indicate that while the

188 CHAPTER 6: The Future for Scanning Helium Microscopy

extended beam lengths are beneficial when developing these more complex optical elements, ultimately the future for neutral helium microscopy lies in more compact beamlines in order to maximise the intensity.

6.1.2 Neutral Atom Microscope (‘NAM’)

The only other instrument besides the Mark I and II SHeM systems capable of exploiting neutral helium as a probe particle to generate a reflection mode image features a novel implementation of a free-jet source and subsequent optic element. Capable of producing images with a resolution better than half a micron [86], ‘neutral atom microscopy’ (‘NAM’) combines the skimmer and a downstream pinhole (as might be found in a SHeM) into a single aperture, allowing the source and sample to be brought as close as possible. As can be seen in the schematic of the instrument geometry in Figure 6.3, the ‘conical aperture holder’ sits in the position normally reserved for the skimmer, but inverted in direction along the optical axis. In this way, the intensity of the spot produced on the sample (placed at normal incidence relative to the beam) is maximised. The count rates are sufficiently high that the pinhole diameter can be brought down to as low as a few hundred nanometres, affording the technique its impressive resolution (the most recent iteration [87] reports a value as low as 315 nanometres).

Figure 6.3 Schematic view of the instrument geometry for the neutral atom microscope. In the most recent iterations, the nozzle is held between 300 and 600 microns from the pinhole aperture while the working distance is typically between 10 and 50 microns. Together, the minimisation of the distance from source to sample enables the instrument to generate a large helium flux incident on the sample surface. Image courtesy of Witham et. al. [87].

189 CHAPTER 6: The Future for Scanning Helium Microscopy

The large count rate allows for superior collection statistics and high pixel counts while still maintaining reasonable scan times (~10 hours for publication quality images). Combined with the sub-micron resolution, the instrument produces excellent images of topologically contrasting samples (see Figure 6.4 for examples) while ensuring no potential for damage through the properties of the neutral probe particle. The design of the instrument represents a significant deviation from a standard arrangement of elements for an atomic beam system, with the changes having specific implications for the nature of the produced images.

Figure 6.4 Micrographs as produced by the Neutral Atom Microscope (‘NAM’). (a) 50nm thick crumpled gold film overlaid on mica background (scale bar 40um). (b) Multilayer graphene (scale bar 50um). (c) Crocosmia pollen grain (scale bar 30um). Images courtesy of Witham et. al. [87]

The purpose of the skimmer in traditional atomic and molecular beam sources is to isolate the centreline of the expansion while reducing interference from the remainder of the beam. The backscattering of atoms into the expansion represents a considerable source of attenuation to the beam and hence a loss of signal – as such, the aerodynamics of the skimmer profile has received considerable attention [34, 95, 133-135, 156, 157]. The inversion of the skimmer in the NAM minimises the distance from pinhole to sample while also allowing for an extended detector aperture, illustrating the core rational of maximising available signal. However, this focus does come at some cost – a significant portion of the atoms that are not part

190 CHAPTER 6: The Future for Scanning Helium Microscopy

of the centreline of the expansions passing through the final aperture must be directed back into the expansion, interfering with the free-jet beam by generating a localised concentration of helium partial pressure.

Much as was observed in the different iterations of the SHeM, a secondary effusive beam forms (termed a ‘spray flow’ by the authors of [87]) and then competes with the primary beam. The strength of this secondary beam however will be much higher than that of the SHeM, owing to the fact that the entire free-jet expansion shock structure will be contained within the conical holder. For the higher resolution version of the instrument detailed in [152] (1.4 bar beam stagnation pressure expanding through a 2 micron nozzle), flow calculations such as those given by Miller [34] yield ca. 6 × 1016 atoms/second (2.5 × 10-3 mBar.litres/second) through the nozzle into the restricted volume. The conductance from the holder back into the source chamber (see calculations for a tapered pipe of circular cross-section in Roth [124]) is only of the order of 6 × 10-2 litres/second, and so the local pressure within the cone is approximately 4 × 10-2 mBar. Calculations of the Mach disc location using the background pressure and relevant beam parameters place it within the nozzle-aperture separation, meaning that the instrument is no longer sampling from the zone of silence of the expansion.

The design then finds itself in an interesting position concerning improvement to the resolution and intensity. The authors note that if the source nozzle was retracted too far from the aperture diffuse images were produced, and count rates would suffer at higher stagnation pressures [152]. In the most recent iteration, a nozzle-aperture separation of 300-600 microns and a working distance of 10-50 microns resulted in images with a resolution of less than 300 microns. However, the difficulty in pushing these separations to ever smaller values and the effects of the restricted pumping mean that further advancements in such a manner are limited. The low beam stagnation pressure, along with sampling the expansion past the Mach disc, means that the produced beam incident on the sample surface will have a very broad energy spread [30, 34, 94, 95]. Investigations involving chemical contrast ideally require a characterised probe particle energy with low velocity spread so as to maximise the available contrast between materials due to interactions with specific vibrations. With the size of the contrast due to chemical effects already small, the diminished degree of monochromaticity would suggest that the instrument is not well suited to such studies.

191 CHAPTER 6: The Future for Scanning Helium Microscopy

The other main difference to the other systems capable of producing neutral helium micrographs is the scattering geometry, namely the beam striking the sample at normal incidence with a detector slotting in from the side. In order to compare the effect the geometry has on the produced images, we return to the relationship defined for topological contrast. Assuming diffuse elastic scattering from a microscopically rough specimen, equation 5.1 can be used to determine the contrast between two planes inclined by ±δ from a mean plane whose normal is itself inclined at an angle of θ with respect to the detector direction. As judged from the schematic in [152] (Figure 6.3), the detector aperture for the NAM will span perhaps some 50 – 90 degrees with respect to the sample normal. Both the Mark II SHeM and the NAM θ ranges have been added to Figure 5.10 to illustrate the difference between scattering geometry of the instruments.

Figure 6.5 Plot of the magnitude of the topological contrast as given by equation 5.1 for a range of values of θ and δ. Note that when the condition that θ + δ < 90o is broken (i.e.: the detector line-of-sight is occluded - top right half of the plot), the contrast has been set to zero for readability. The region between the blue dot-dashed lines indicate the range of θ values for the Mark II SHeM, while that between the red dashed lines show the equivalent for the NAM.

The net result of the necessary detector placement for the NAM is a constraint to the aspect ratios of the samples it images – samples with large mean plane deviations will quickly have portions of the surface cast into pure shadow. This outcome is also reinforced by the minimal working distance crucial to the success

192 CHAPTER 6: The Future for Scanning Helium Microscopy

of the design (due to the divergence of the beam from the pinhole). Samples with high aspect ratios carry with them the potential for damage to the pinhole through collision during scanning, as well as the ability to block the detector aperture. Improvements to the resolution for the instrument require shrinking the pinhole and the working distance down further, as well as the size of the pinhole membrane – if not of the order of the working distance, there is a risk of promoting multiple reflections or reducing intensity through detector occlusion.

As such, the neutral atom microscope design is then one that will excel at measuring very small topological deviations on flat samples – that is, surface roughness effects. The impressive resolution and high count rates despite the current limitations on optics and detector technology will allow for small changes in the surface asperity size – even those below the resolution of the instrument – to show up quite vividly. Soft materials adverse to roughness characterisation by means of a stylus (AFM, profilometry, etc.) or exposure to light (optical profilometry methods) would then be ideal candidates for samples. The design also benefits from the minimal amount of vacuum equipment needed to set up a viable instrument. One could see the development of a small, cost effective version of the NAM which, while only able to take advantage of topological contrast, would find significant use.

6.2 Contrast Specific Requirements

6.2.1 Topological Samples

Topological contrast will always be the most dominant mechanism available to the SHeM, and will favour samples which will primarily reply on the non-destructive and inert nature of the probe particle. Biological materials, organic electronics, devices with electric and magnetic strayfields and ultrathin surface coatings are all examples of samples for whom gentle imaging with no required sample preparation would be of great importance. The energy of a thermal neutral helium atom is already low enough to ensure the desired behaviour, and so considerations for a design which seeks to exploit topology primarily revolve around the quality of the imaging, namely the scan speed and the resolution available. If an instrument were able to provide short imaging times with a resolution somewhere of the order of 100 nm and no surface preparation, it would make for a compelling complement to other surface analysis techniques, one that is eminently possible based on the experience with the two systems detailed in the current work.

193 CHAPTER 6: The Future for Scanning Helium Microscopy

The Mark II SHeM is still a developmental instrument, able to have quite serious changes to the components in the process of refining our understanding of the technique. While the Mark II represented a large increase in the available count rates as compared to the previous generation, collecting images with a good signal-to-noise ratio still required imaging for many hours. Increasing the collected signal allows for reductions in the scan time (less dwell per pixel) or an improvement to the resolution. A more finalised instrument design has no need to be as large, or even as expensive; considerations which although not directly relevant to the science are critical to the adoption of the technique as a whole.

In the vein of the modern commercial SEM, an ideal instrument is desktop based, meaning the largest components need to be reduced in size. Bringing the source chamber down to a more manageable dimension turns out to be quite a straightforward matter. Manipulation of the source position within the chamber would be redundant for a fixed nozzle diameter and skimmer size once optimised for intensity (and perhaps resolution). As such, use of the bulky 3-axis manipulator currently used for the Mark II SHeM source would be entirely avoidable. Furthermore, as topological contrast is not dependent on the temperature of the beam, implementing cooling into such a design is not necessary. The main consideration on source size is then the large and expensive turbomolecular pump needed such to handle the helium throughput. Based on recent work by Barr [32], it has been shown that for a source operating in the intermediate regime such as the Newcastle source, bringing the pump rate down significantly does not drastically damage the quality of the image produced (see Figure 6.6 below). In a scans over a silicon knife-edge, the signal-to-background ratio was calculated as a function of the source turbo pump rate by spinning down the pump from full speed (~1500 L/s). It was found that the signal-to-background ratio dropped from 0.471 at full speed to 0.391 at a speed of ~570 L/s, a pumping speed achievable with modern 8” turbo pumps. It should be noted that the described experiment was conducted with an original Mark II pinhole plate (prior to the improvements described in Section 5.2.2), and so the signal-to-background ratios are approximately half of what is currently achievable on the system. A loss of ~20% of the signal-to-background ratio is quite an acceptable concession for a significant reduction in the size of the source turbo.

194 CHAPTER 6: The Future for Scanning Helium Microscopy

Figure 6.6 Plot of the signal-to-background ratio as a function of source chamber pump rate (controlled via the pump rotation speed). Signal-to-background ratios determined from reflected intensity on and off a silicon wafer. Data courtesy of Matthew Barr [32].

In terms of reductions in size, the sample chamber will follow a similar path to the proposed source changes. The footprint of the sample mount – even the larger 3D capable mount designed by Myles [123] – is far smaller than the chamber used in the Mark II design. The turbomolecular pump diameter is the primary driver for the dimensions, but performing the same experiment detailed previously to determine the signal-to-background ratio as a function of sample chamber pump speed (Figure 6.7) shows potential for further reductions in size.

Figure 6.7 Plot of the signal-to-background ratio as a function of sample chamber pump rate for a section of flat silicon oxide surface. Silicon chip was mounted above a hole in a sample slide, with the background taken as the count rate obtained from the hole. Note that the experiment utilised an original Mark II pinhole plate, resulting in sub-optimal signal-to-background ratios.

195 CHAPTER 6: The Future for Scanning Helium Microscopy

A drop of perhaps 10% in signal-to-background ratio – again noting that the experiment was performed with an original pinhole plate – puts the pump speed in the range of a 6” turbomolecular pump, and thus a smaller chamber. Considering the success of the sample chamber door as implemented in the Mark II design, a future sample chamber would likely become very similar to those in modern commercial SEM systems, allowing for sample exchange to be carried out with a minimum of effort and imaging able to be started within 10 minutes.

As noted for both iterations of SHeM described in this thesis, the differential pump rate is critical in ensuring the size of the secondary effusive beam is kept to a minimum. Increasing the size of the differential stage turbomolecular pump slightly would then provide a significant enhancement in image quality (similar to the improvements seen with pinhole plate alterations for the Mark II SHeM). Moving towards designs closer in layout to HAS systems, whereby the source and first differential stage are compactly built into a single box chamber separated by an internal wall, should allow for this larger pump while still reducing the size of the instrument. Considering the desired alterations to each of the main chambers, one can easily envision a topologically focused SHeM that succeeds as a desktop instrument.

In addition to reducing the size and complexity of the instrument, the critical aspect to address for topological contrast is the traditional issue of all neutral helium techniques: available intensity. Higher count rates will allow for enhancement to many different aspects of the system: lower noise present in images, faster imaging, higher resolutions, and even the option to reduce the effect of the effusive beam by adding in further differential pumping prior to the pinhole. Of particular importance for the SHeM going forward will be the balance between helium signal and the resolution of the instrument as dictated by the size of the pinhole. Experiments detailed in Section 5.4 have shown that reductions in the pinhole size affect the detected intensity geometrically, and as such smaller spot sizes present a significant loss in signal. If the resolution is also tied to the virtual source size as current experiments seem to indicate, then smaller skimmers will also be necessary, compounding the intensity issue.

Any improvements to the output of a free-jet helium beam source are likely to be minimal, with the designs as utilised in the Mark I and Mark II instruments the result of decades of research and optimisation. There has been some recent interest in how the shape of the nozzle used affects the centreline intensity of the produced expansion. Work by Even [158] shows moving from a sonic nozzle (such as

196 CHAPTER 6: The Future for Scanning Helium Microscopy

employed in the Buckland style sources employed in both generations of SHeM) to a parabolic nozzle sharpens the angular distribution of the free-jet expansion, resulting in an approximately order of magnitude higher number density along the central axis. Further testing would be required before such a switch can be confirmed as a boost to SHeM performance, as it is currently unclear how such a change affects the virtual source size, and its role in instrument resolution.

In terms of optics, the simple pinhole used in the SHeM represent a delicate balance between resolution and intensity. Moving to alternate optic elements may improve the situation, but significant amounts of work are required before they could be implemented in a SHeM. Atomic mirrors for neutral helium are still under development, with new material choices seeking to overcome the issues of expense and cleanliness that dogged prior designs. Designs based around graphene surfaces [65], more recently as grown on thin metal crystal substrates to aid in the bending required to shape the mirror [61], report helium reflectivities of approximately 20%, but the produced spot sizes are still larger than currently achievable through use of a pinhole. Diffractive focusing of the helium beam via zoneplates [23, 57, 58, 141, 153, 159, 160] or atom sieves [161] has seen steady progress over the last decade, with current designs capable of producing a spot size of less than 4 microns. Fabrication methods are improving with advances in lithography (reducing the potential cost of the delicate, free-standing silicon nitride structures), but their use is currently limited by the broad background created by additional diffraction orders, leading to restrictive signal-to-background levels. Implementation of order sorting apertures and further optimisation should see these focusing elements become viable in the near future, but alternate patterns may be required in order to account for the short distance between the SHeM beam source and optical element, or else suffer from the effects of aberrations in the produced helium spot.

An alternative optical element may offer a solution somewhere between the intensity/resolution conflict of a simple pinhole, and the complexity of a zoneplate. Coded apertures are a form of computational optic [162, 163] consisting of an array of openings (often using squares as the basis element), originally employed in high energy astronomy [164, 165]. The flux through the optical element is significantly increased due to the multiple pinholes present, with the recorded intensity a complex overlap of the image produced by each opening. By then deconvolving out the mask used, an image of the sample can be recovered. The resolution of such an optical element is normally attributed to the sharpness of the edges in the

197 CHAPTER 6: The Future for Scanning Helium Microscopy

mask, the signal-to-noise ratio of the collected image, and the quality of the deconvolution with the point spread function of the mask. The photon sieve constructed by Holst et. al. [161] can be considered a form of coded aperture, and the principles of the design are highly amenable to the current generation of SHeM. It should be fairly straight-forward to design an appropriate mask pattern that can be milled into a silicon nitride membrane in the same manner as the current pinholes. Once complete, the membrane can be affixed to a standard pinhole plate for testing within the system.

While changes to the beam source and neutral atom optics are possible, the most likely source of a significant improvement to the microscope intensity will be due to new detector technology. As discussed in Chapter 1, all of the properties which make neutral helium atoms the ideal probe also ensure that it very difficult to detect through the usual ionisation methods. While continual refinements are made to commercial quadrupoles – as seen in the three times sensitivity increase in the Hiden mass spectrometer design between the Mark I and Mark II instruments – ideally something a little more revolutionary is possible. The continuing work on the solenoidal ion source has been motivated by a particular interest in HAS and SHeM applications, and offers the potential for a 104 – 105 times increase in detector sensitivity as compared to a commercial quadrupole. The most current versions in the literature quote a sensitivity of approximately 0.38 A/mbar for helium [74], and so their incorporation into the microscope should see an orders of magnitude increase in the recorded count rate. Care will need to be taken with respect to the rise time of any such detector – as seen with the Mark I and Mark II instruments, should the stagnation volume not reach equilibrium quickly enough, the result will be dragging artefacts across sharp features in the produced micrographs. The internal volume of a solenoidal ion source is much larger than that for a commercial quadrupole, and so ample pumping and effective ion extraction will be essential to achieving the desired performance.

198 CHAPTER 6: The Future for Scanning Helium Microscopy

6.2.2 Chemical Contrasts

Chemical contrast relies on energy exchanges with surface phonons or adsorbates. With the relative frequency of inelastic scattering events (as compared to elastic ones), improving the count rate is an even greater priority for a microscope that seeks to exploit them as a contrast mechanism. The strength of the contrast observed for the metallic layers on silicon covered in Chapter 5 necessitated much longer dwell times than for any topological samples; a practical microscope will need to collect images in a more brisk fashion. The count rate issue extends beyond speed and image quality, as the contrast between different materials may be enhanced by reducing the size of the detector aperture. If a future instrument maintains the scattering geometry of the Mark I and Mark II systems, then chemical contrast will continue to revolve around how much signal is diverted out of the specular channel. A smaller detector aperture will provide superior angular resolution, and thus be better equipped to uncover small variations in the scattered intensity. The 1mm diameter detector aperture used in the Mark II design subtends an angle of approximately 20.5º from a sample plane at specular, and so significant improvement in this regard will be possible. Assuming a geometric reduction in count rate as seen with the experiments involving smaller pinholes, improving the angular resolution of the instrument will cause a significant loss in intensity. Any of the various possible count rate upgrade paths discussed with reference to topological contrast should be able to help counter such losses.

Beyond issues of count rate, there are several possible design features which would allow the contrast method to reach its full imaging potential:

 Helium Probe Energy: To tailor the probe energy, cooling of the beam stagnation volume is a core requirement of a chemical contrast orientated instrument. The counterflow method employed in the Mark II system proved quite effective as a low cost alternative to more traditional chilling system, but several improvements in performance would benefit a dedicated instrument. In particular, expanding the possible temperature range of the source [166], or more precisely controlling the temperature stability [167], would be of great benefit to long scan times.

 Surface Cleaning: Chemical contrast will be the most useful when the incident helium atoms are not striking a surface covered with adsorbates, but instead interacting with the atoms of the material in question directly. The normal overlayers on a surface (most commonly water, oxygen, and carbonaceous layers) will reduce the likelihood of scattering pathways which contribute to these contrast mechanisms. Such concerns are nothing new to the field of surface science, with many systems

199 CHAPTER 6: The Future for Scanning Helium Microscopy

requiring some means to strip away impurities. For example, in HAS studies it is well known that the neutral helium is so surface sensitive that individual hydrogen atoms adsorbed onto a surface will distort the electronic corrugation sufficiently to cause significant scattering outside the expected diffraction pattern. While useful for adsorbate studies, when a study is attempting to understand the nature of the materials beneath the adsorbates this is simply a loss in useful signal. Multiple methods of cleaning a surface are possible, and are generally matched to the sample material in question:

o For single crystal surfaces, direct heating of the sample to high temperatures in order to drive off any potential contaminants is common. A process known as flashing [168], the technique is generally only suitable for homogeneous materials which do not suffer damage under the heating cycles, thus limiting its usefulness for SHeM.

o Sputter cleaning is a process wherein an ion beam, commonly argon, is directed at the sample surface to physically remove surface layers [169]. While more gentle than flashing, one must be careful with issues of preferential etching of certain atomic species, heating, and contamination from the ion beam. Considering the strengths of SHeM at imaging delicate materials, its use would be restricted to a subset of the samples entered into a system.

o A more recent addition for cleaning, and one that matches well to more fragile samples, is a cluster ion beam source [170]. A free-jet expansion (often argon due to its inert nature) is produced with the conditions controlled in such a way to encourage the formation of large clusters of atoms. Subsequent ionisation of these clusters produces projectiles with a very large mass but an almost negligible charge. Electrostatic manipulation of the ion clusters then forms a beam to be accelerated towards the sample surface. The impact of the clusters removes surface asperities and adsorbates at a much lower interaction energy than the previously described methods, and thus acts to clean and smooth. The sensitive nature of the cleaning for a cluster ion beam would then push it to the front of the list of likely candidates.

 Ultra-high vacuum (UHV): Surface cleaning would be useless unless that surface is able to remain contaminate free for an extended period. As such, the vacuum requirements for the sample chamber become more stringent than those used for the Mark I and Mark II systems. For a SHeM, the desired period of time would match to the length of whatever scan is to be conducted – likely meaning a chamber pressure in the range of 10-10 mBar or better (equating to a monolayer formation time of greater than an hour). The lower pressure will also mean more careful consideration of sample choice due to the issue of outgassing.

200 CHAPTER 6: The Future for Scanning Helium Microscopy

 Sample Temperature Control / Electrical connections: The experiments conducted using the Mark II SHeM involving chemical contrast revolved around using material properties (atomic mass and Debye temperature) and control over the helium probe energy through stagnation temperature. Inspection of the formulation of the DWF (equation 1.5) shows it should also be possible to control the contrast by changing the temperature of the sample under investigation. Furthermore, should there be a temperature gradient in the sample then this would also show up as contrast under the helium beam. An example of where this effect may be useful is in the study of devices with electrical circuits – the nature of the helium probe means that the devices could be active during imaging for a unique perspective into their operation. Even organic photovoltaics could be scanned whilst under illumination – a powerful combination of the delicate imaging possible with the SHeM and its multiple contrast methods. The addition of direct sample heating and/or electrical connections then opens up a wide range new imaging opportunities.

6.2.3 Diffractive Contrast

Sample systems exhibiting diffractive contrast will be very similar those often found in HAS studies – surface coatings, thin film growth, and phonon dispersion measurements. Special requirements to enhance the available contrast are, for the most part, as discussed for chemical contrast: higher count rates, sample cleaning to produce well-ordered surfaces, and sample chamber pressures in the UHV range to preserve the surface. The 3D mount discussed in Section 5.1.1 (or something similar) would be a critical addition, allowing the sample to be tilted relative to the incident beam. HAS systems typically go a step beyond this, and are able to sweep the detector around the sample. Considering the limited space available in the current designs, such an addition to SHeM would require a significant redesign and likely a loss in signal at the detector due to a longer beamline.

However, such a change may be inevitable due to the biggest concern for diffractive effects: the angular resolution of the detector. To be able to sufficiently differentiate between different order diffraction peaks, HAS systems are able to resolve down to at least 0.5º - a concern more critical than even the count rate, and a major driver for the extended beamline lengths. As mentioned above, the Mark II instrument currently has an angular resolution of approximately 20.5º for a sample at specular. Even if the detector aperture were replaced with a silicon nitride membrane with a 5 micron aperture (i.e.: our standard pinhole), the resulting resolution would only drop to approximately 0.8º, and calculations of the available

201 CHAPTER 6: The Future for Scanning Helium Microscopy

intensity (using a geometric reduction) show the available signal drops to zero. Even more conservative aperture reductions – likely able to provide some indication of the contrast method - are not promising in terms of count rate: a 100 micron aperture yields an angular resolution of ~ 2.7 º and an intensity of only 1% of that of the original.

It is then apparent that improvements to the available signal are necessary before diffractive contrast is practical. Should higher count rates become available, a move back towards a more HAS-like system (longer beamline, more differential stages) may be required to specialise in this contrast mechanism. It may even be that spatially resolved diffraction is an option built in to dedicated HAS system, as opposed to an imaging mode for a SHeM.

6.3 Conclusions

The concept of a microscopy technique employing neutral helium as imagined over 20 years ago has come to pass. The confirmation that imaging with helium atoms is not only feasible but carries with it all of the proposed benefits shows great promise for the future of the technique. While still an early area of research, it is not hard to imagine SHeM becoming a standard addition to the toolset of surface science.

The work presented in this thesis has demonstrated not one but two separate instruments, with the latter building on the success of the first to push the limits of the technique. Direct observation of both topological and chemical contrasts demonstrates the power of the SHeM for a wide range of samples which conventional microscopies would struggle to image. The avenues of research opened up through the development of these instruments – image formation, effusive beam effects, the nature of atom beam optics, and fundamental studies into the nature of the available contrast – will all go forward to enhance the strengths of any future designs. Combined with a few choice enhancements to neutral atom optics and detection, the next generation of instruments will truly be capable of providing valuable insight for the broader scientific community.

202 References

REFERENCES

1. Feynman, R.P. and A. Zee, QED: The Strange Theory of Light and Matter. STU - Student edition ed. 1985: Princeton University Press. 2. Lambert, L. and T. Mulvey, Ernst Ruska (1906–1988), Designer Extraordinaire of the Electron Microscope: A Memoir, in Advances in Imaging and Electron Physics, P.W. Hawkes, Editor. 1996, Elsevier. p. 2-62. 3. Kisielowski, C., et al., Detection of Single Atoms and Buried Defects in Three Dimensions by Aberration-Corrected Electron Microscope with 0.5-Å Information Limit. Microscopy and Microanalysis, 2008. 14(5): p. 469-477. 4. Notte, J. and J. Huang, The Helium Ion Microscope, in Helium Ion Microscopy, G. Hlawacek and A. Gölzhäuser, Editors. 2016, Springer International Publishing: Cham. p. 3-30. 5. Liu, D., et al., Demonstration of a novel focusing small-angle neutron scattering instrument equipped with axisymmetric mirrors. Nature Communications, 2013. 4: p. 2556. 6. Pantano, C.G. and T.E. Madey, Electron beam damage in Auger electron spectroscopy. Applications of Surface Science, 1981. 7(1): p. 115-141. 7. Egerton, R.F., Control of radiation damage in the TEM. Ultramicroscopy, 2013. 127: p. 100-108. 8. Egerton, R.F., P. Li, and M. Malac, Radiation damage in the TEM and SEM. Micron, 2004. 35(6): p. 399-409. 9. Horowitz, G., Organic Transistors, in Organic Electronics. 2006, Wiley-VCH Verlag GmbH & Co. KGaA. p. 1-32. 10. Grubb, D.T., Radiation damage and electron microscopy of organic polymers. Journal of Materials Science, 1974. 9(10): p. 1715-1736. 11. Kitching and Donald, Beam damage of polypropylene in the environmental scanning electron microscope: an FTIR study. Journal of Microscopy, 1998. 190(3): p. 357-365. 12. Isaacson, M., D. Johnson, and A.V. Crewe, Electron Beam Excitation and Damage of Biological Molecules; Its Implications for Specimen Damage in Electron Microscopy. Radiation Research, 1973. 55(2): p. 205-224. 13. Butler, J.H., et al., Low-voltage scanning electron microscopy of polymers. Polymer, 1995. 36(9): p. 1781-1790. 14. Sezen, M., et al., An investigation on focused electron/ion beam induced degradation mechanisms of conjugated polymers. Physical Chemistry Chemical Physics, 2011. 13(45): p. 20235-20240. 15. Myles, T., et al., Atom Microscopy - Imaging with a deft touch, in Australian Physics. 2017, AIP: Melbourne, Victoria, Australia.

References

16. Traeger, F., Helium Atom Scattering from Oxide Surfaces. ChemPhysChem, 2006. 7(5): p. 1006-1013. 17. Ichimura, S. and R. Shimizu, Observation of electron beam damage in thin‐film SiO2 on Si with scanning Auger electron microscope. Journal of Applied Physics, 1979. 50(9): p. 6020-6022. 18. Goldstein, J.I., et al., Scanning Electron Microscopy and X-ray Microanalysis: Third Edition. 2003, Springer US: Boston, MA. 19. MacLaren, D.A., et al., Focusing Elements and Design Considerations for a Scanning Helium Microscope (SHeM). Surface Review and Letters, 2003. 10(02n03): p. 249-255. 20. Krivanek, O.L., et al., Vibrational spectroscopy in the electron microscope. Nature, 2014. 514: p. 209. 21. Engel, T. and K.H. Rieder, Structural studies of surfaces with atomic and molecular beam diffraction, in Structural Studies of Surfaces, K. Heinz, et al., Editors. 1982, Springer Berlin Heidelberg: Berlin, Heidelberg. p. 55-180. 22. Cavendish Laboratories SMF Surface Physics Group. A Brief Introduction to Helium Atom Scattering. [cited 2016 August 7th]; Available from: https://web.archive.org/web/20120207085443/http://www- sp.phy.cam.ac.uk/research/HASoverview.php3. 23. Doak, R.B., et al., Towards Realization of an Atomic de Broglie Microscope: Helium Atom Focusing Using Fresnel Zone Plates. Physical Review Letters, 1999. 83(21): p. 4229-4232. 24. MacLaren, D.A., Development of a Single Crystal Mirror for Scanning Helium Microscopy. 2002, University of Cambridge, England. 25. Daniel, F. and R. Karl-Heinz, Atomic beam diffraction from solid surfaces. Reports on Progress in Physics, 1998. 61(12): p. 1575. 26. B. Poelsema and G. Comsa, Scattering of Thermal Energy Atoms. Springer Tracts in Modern Physics. Vol. 115. 1989: Springer-Verlag, Berlin. 27. Manson, J.R., Inelastic scattering from surfaces. Physical Review B, 1991. 43(9): p. 6924-6937. 28. Gumhalter, B., Single- and multiphonon atom–surface scattering in the quantum regime. Physics Reports, 2001. 351(1): p. 1-159. 29. Yinnon, A.T., et al., He scattering from random adsorbates, disordered compact islands, and fractal submonolayers: Intensity manifestations of surface disorder. The Journal of Chemical Physics, 1997. 106(10): p. 4228-4242. 30. Helium Atom Scattering from Surfaces. Springer Series in Surface Sciences, ed. E. Hulpke. Vol. 27. 1992, Berlin Heidelberg: Springer-Verlag. 31. MacLaren, D.A. and W. Allison, Microscopy with atomic beams: contrast in a scanning helium microscope. Institute of Physics Conference Series, 2004. 179: p. 383-388. 32. Matthew G. Barr, Imaging with Atoms: Aspects of Scanning Helium Microscopy. 2015, University of Newcastle, Australia.

204 References

33. Toennies, J.P., Scattering of molecular beams from surfaces. Applied physics, 1974. 3(2): p. 91-114. 34. Miller, D.R., in Atomic and Molecular Beam Methods, G. Scoles, Editor. 1988, Oxford University Press. 35. Breedlove, J.R. and G.T. Trammell, Molecular Microscopy: Fundamental Limitations. Science, 1970. 170(3964). 36. Handbook of Charged Particle Optics. 2nd Edition ed, ed. J. Orloff. 1997: CRC Press. 37. Michelson, A.A., Studies in optics. 1995: Courier Corporation. 38. Benedek, G., et al., Unveiling mode-selected electron-phonon interactions in metal films by helium atom scattering. Physical Chemistry Chemical Physics, 2014. 16(16): p. 7159-7172. 39. Seiler, H., Secondary electron emission in the scanning electron microscope. Journal of Applied Physics, 1983. 54(11): p. R1-R18. 40. Toennies, J.P., Helium atom scattering: a gentle and uniquely sensitive probe of surface structure and dynamics. Journal of Physics: Condensed Matter, 1993. 5(33A): p. A25. 41. Doak, R.B., in Atomic and Molecular Beam Methods, G. Scoles, Editor. 1992, Oxford University Press. 42. Graham, A.P. and J.P. Toennies, Determination of the lateral potential energy surface of single adsorbed atoms and molecules on single crystal surfaces using helium atom scattering. Surface Science, 1999. 427-428: p. 1-10. 43. Chis, V., et al., Large Surface Charge Density Oscillations Induced by Subsurface Phonon Resonances. Physical Review Letters, 2008. 101(20): p. 206102. 44. Benedek, G., et al., Atoms riding Rayleigh waves. Journal of Physics: Condensed Matter, 2010. 22(30): p. 304016. 45. Comsa, G., The applicability of the Debye-Waller factor to the He scattering from metal surfaces. Journal of Physics C: Solid State Physics, 1973. 6(17): p. 2648. 46. Manson, J.R., G. Benedek, and S. Miret-Artés, Electron–Phonon Coupling Strength at Metal Surfaces Directly Determined from the Helium Atom Scattering Debye–Waller Factor. The Journal of Physical Chemistry Letters, 2016. 7(6): p. 1016-1021. 47. Kühnle, A., Self-assembly of organic molecules at metal surfaces. Current Opinion in Colloid & Interface Science, 2009. 14(2): p. 157-168. 48. Dastoor, P.C. and W. Allison, Geometric contrast mechanisms in helium atom scattering: The growth of the Fe/Cu(100) system. Physical Review B, 2003. 67(24): p. 245403. 49. van Bavel, S.S., et al., P3HT/PCBM Bulk Heterojunction Solar Cells: Impact of Blend Composition and 3D Morphology on Device Performance. Advanced Functional Materials, 2010. 20(9): p. 1458-1463. 50. Morse, M.D., Supersonic beam sources. Experimental methods in the physical sciences, 1996. 29: p. 21-47.

205 References

51. Pauly, H., Atom, Molecule, and Cluster Beams I (Basic Theory, Production and Detection of Thermal Energy Beams). Springer Series on Atomic, Optical, and Plasma Physics. 2000, Berlin Heidelberg: Springer-Verlag 52. Halwidl, D., Development of an effusive molecular beam apparatus. 2016: Springer. 53. Rugamas, F., et al., Angular profiles of molecular beams from effusive tube sources: I. Experiment. Measurement Science and Technology, 2000. 11(12): p. 1750. 54. Kantrowitz, A. and J. Grey, A High Intensity Source for the Molecular Beam. Part I. Theoretical. Review of Scientific Instruments, 1951. 22(5): p. 328-332. 55. Buckland, J.R., et al., A simple nozzle design for high speed-ratio molecular beams. Measurement Science and Technology, 1997. 8(8): p. 933. 56. Eder, S.D., et al., Two Dimensional Imaging of the Virtual Source of a Supersonic Beam: Helium at 125 K. The Journal of Physical Chemistry A, 2014. 118(1): p. 4-12. 57. Reisinger, T., et al., Free-standing silicon-nitride zoneplates for neutral-helium microscopy. Microelectronic Engineering, 2010. 87(5): p. 1011-1014. 58. Koch, M., et al., Imaging with neutral atoms—a new matter‐wave microscope. Journal of microscopy, 2008. 229(1): p. 1-5. 59. Holst, B. and W. Allison, An atom-focusing mirror. Nature, 1997. 390: p. 244. 60. Weeks, A., Si(111) Atom-Optical Mirrors for Scanning Helium Microscopy. 2007, University of Cambridge. 61. Anemone, G., et al., Flexible thin metal crystals as focusing mirrors for neutral atomic beams. Physical Review B, 2017. 95(20): p. 205428. 62. Barredo, D., et al., A Quantum-Stabilized Mirror for Atoms. Advanced Materials, 2008. 20(18): p. 3492-3497. 63. Barredo, D., et al., A high-reflectivity atom-focusing mirror stable at room temperature. Applied Physics Letters, 2010. 96(8): p. 081901. 64. Fladischer, K., et al., An ellipsoidal mirror for focusing neutral atomic and molecular beams. New Journal of Physics, 2010. 12(3): p. 033018. 65. Sutter, P., et al., A high-reflectivity, ambient-stable graphene mirror for neutral atomic and molecular beams. Applied Physics Letters, 2011. 99(21): p. 211907. 66. Doak, R.B., The assessment of field ionization detectors for molecular beam use. Journal of Physics: Condensed Matter, 2004. 16(29): p. S2863. 67. Doak, R.B., et al., Field ionization detection of supersonic molecular beams. Review of Scientific Instruments, 2004. 75(2): p. 405-414. 68. Liao, L., et al., A novel gas sensor based on field ionization from ZnO nanowires: moderate working voltage and high stability. Nanotechnology, 2008. 19(17): p. 175501. 69. Gross, J.H., Mass spectrometry: a textbook. 2006: Springer Science & Business Media.

206 References

70. De Hoffmann, E. and V. Stroobant, Mass spectrometry: principles and applications. 2007: John Wiley & Sons. 71. Märk, T.D. and G.H. Dunn, Electron impact ionization. 2013: Springer Science & Business Media. 72. Rejoub, R., B.G. Lindsay, and R.F. Stebbings, Determination of the absolute partial and total cross sections for electron-impact ionization of the rare gases. Physical Review A, 2002. 65(4): p. 042713. 73. DeKieviet, M., et al., Design and performance of a highly efficient mass spectrometer for molecular beams. Review of Scientific Instruments, 2000. 71(5): p. 2015-2018. 74. Chisnall, D., A High Sensitivity Detector for Helium Atom Scattering. 2012, University of Cambridge. 75. Mainfray, G. and G. Manus, Multiphoton ionization of atoms. Reports on Progress in Physics, 1991. 54(10): p. 1333. 76. Dass, C., Modes of Ionization, in Fundamentals of Contemporary Mass Spectrometry. 2007, John Wiley & Sons, Inc. p. 15-65. 77. Pauly, H., Atom, Molecule, and Cluster Beams II: Cluster Beams, Fast and Slow Beams, Accessory Equipment and Applications. Vol. 32. 2013: Springer Science & Business Media. 78. Murphy, E.L. and R.H. Good, Thermionic Emission, Field Emission, and the Transition Region. Physical Review, 1956. 102(6): p. 1464-1473. 79. Oppenheimer, J.R., Three Notes on the Quantum Theory of Aperiodic Effects. Physical Review, 1928. 31(1): p. 66-81. 80. O'Donnell, K., Field Ionisation Detection for Atom Microscopy. 2009, University of Newcastle, Australia. 81. Miller, M.K., et al., Atom probe field ion microscopy. 1996. 82. Tsong, T.T., Atom-probe field ion microscopy: field ion emission, and surfaces and interfaces at atomic resolution. 2005: Cambridge University Press. 83. Riley, D.J., et al., Helium Detection via Field Ionization from Carbon Nanotubes. Nano Letters, 2003. 3(10): p. 1455-1458. 84. O'Donnell, K.M., et al., Field ionization detection of helium using a planar array of carbon nanotubes. Physical Review B, 2012. 85(11): p. 113404. 85. R. B. Sadeghian and M. S. Islam, Ultralow-voltage field-ionization discharge on whiskered silicon nanowires for gas-sensing applications. Nature Materials, 2011. 10: p. 135. 86. Witham, P. and E. Sánchez, A simple approach to neutral atom microscopy. Review of Scientific Instruments, 2011. 82(10): p. 103705. 87. Witham, P. and E. Sánchez, Exploring neutral atom microscopy. Crystal Research and Technology, 2014. 49(9): p. 690-698. 88. Economou, N.P., J.A. Notte, and W.B. Thompson, The history and development of the helium ion microscope. Scanning, 2012. 34(2): p. 83-89. 89. Smith, S.W., The scientist and engineer's guide to digital signal processing. 1997.

207 References

90. OLYMPUS CORPORATION. Microscopy Resource Center. [cited 2015 April 29th]; Available from: https://www.olympus-lifescience.com/en/microscope- resource/. 91. Greenwood, J., The correct and incorrect generation of a cosine distribution of scattered particles for Monte-Carlo modelling of vacuum systems. Vacuum, 2002. 67(2): p. 217-222. 92. Chambers, A., Modern vacuum physics. 2004: CRC Press. 93. Hiden Analytical Ltd. 2011. 94. Campargue, R., Progress in overexpanded supersonic jets and skimmed molecular beams in free-jet zones of silence. The Journal of Physical Chemistry, 1984. 88(20): p. 4466-4474. 95. Beijerinck, H.C.W., et al., Campargue-type supersonic beam sources: Absolute intensities, skimmer transmission and scaling laws for mono-atomic gases He, Ne and Ar. Chemical Physics, 1985. 96(1): p. 153-173. 96. W. Allison and A. Jardine. 2011. 97. Barr, M., et al., A desktop supersonic free-jet beam source for a scanning helium microscope (SHeM). Measurement Science and Technology, 2012. 23(10): p. 105901. 98. Scientific Volume Imaging. Signal-to-Noise Ratio. [cited 2015 April 29th]; Available from: https://svi.nl/SignalToNoiseRatio. 99. Carl Zeiss Microscopy. Education in Microscopy and Digital Imaging. Available from: http://zeiss-campus.magnet.fsu.edu/articles/basics/digitalimaging.html. 100. Fouquet, P., et al., Thermal energy He3 spin-echo spectrometer for ultrahigh resolution surface dynamics measurements. Review of Scientific Instruments, 2005. 76(5): p. 053109. 101. Beijerinck, H.C.W. and N.F. Verster, Absolute intensities and perpendicular temperatures of supersonic beams of polyatomic gases. Physica B+C, 1981. 111(2): p. 327-352. 102. Oren, M. and S.K. Nayar, Generalization of Lambert's reflectance model, in Proceedings of the 21st annual conference on Computer graphics and interactive techniques. 1994, ACM. p. 239-246. 103. Nayar, S.K. and M. Oren, Visual appearance of matte surfaces. Science, 1995. 267(5201): p. 1153. 104. Goldstein, J.I., et al., Scanning electron microscopy and X-ray microanalysis. 2017: Springer. 105. Grogan, D.F., et al., Planarizing technique for ion-beam polishing of diamond films. Applied Optics, 1992. 31(10): p. 1483-1487. 106. Sirineni, G.M.R., et al., Reactive ion etching of diamond as a means of enhancing chemically-assisted mechanical polishing efficiency. Diamond and Related Materials, 1997. 6(8): p. 952-958. 107. Kunuku, S., et al., Investigations on Diamond Nanostructuring of Different Morphologies by the Reactive-Ion Etching Process and Their Potential Applications. ACS Applied Materials & Interfaces, 2013. 5(15): p. 7439-7449.

208 References

108. Pawley, J., Handbook of biological confocal microscopy. 2010: Springer Science & Business Media. 109. Agar Scientific. Tin on carbon AGS1937 - universal resolution specimen. [cited 2015 August 11th]; Available from: http://www.agarscientific.com/universal- resolution-tin-on-carbon-5nm-30-181-m-specimen.html. 110. Chen, P., F. Huang, and Y. Ding, Microstructure, deformation and failure of polymer bonded explosives. Journal of Materials Science, 2007. 42(13): p. 5272- 5280. 111. Kern, K., R. David, and G. Comsa, Low‐energy helium nozzle beam. Review of Scientific Instruments, 1985. 56(3): p. 369-372. 112. Ekinci, Y., E.L. Knuth, and J.P. Toennies, A mass and time-of-flight spectroscopy study of the formation of clusters in free-jet expansions of normal D2. The Journal of Chemical Physics, 2006. 125(13): p. 133409. 113. Reisinger, T., et al., Brightness and virtual source size of a supersonic deuterium beam. Physical Review A, 2012. 86(4): p. 043804. 114. Beam Dynamics Inc. Molecular Beam Skimmers (Specs). [cited 2016 March 12th]; Available from: http://www.beamdynamicsinc.com/skimmer_specs.htm. 115. attocube. ECS3030 - nanoprecise stepper positioner for linear, horizontal motion. [cited 2017 November 1st]; Available from: http://attocube.com/attomotion/industrial-line/ecsx3030/. 116. Asadi, M., Beet-sugar handbook. 2006: John Wiley & Sons. 117. Hall, C.E., Scattering Phenomena in Electron Microscope Image Formation. Journal of Applied Physics, 1951. 22(5): p. 655-662. 118. Sheppard, C.J.R. and T. Wilson, Image formation in confocal scanning microscopes. Optik, 1980. 55(4): p. 331-342. 119. Tóbik, J., I. Štich, and K. Terakura, Effect of tip morphology on image formation in noncontact atomic force microscopy: InP(110). Physical Review B, 2001. 63(24): p. 245324. 120. Bell, D.C., Contrast Mechanisms and Image Formation in Helium Ion Microscopy. Microscopy and Microanalysis, 2009. 15(2): p. 147-153. 121. Reimer, L., Scanning electron microscopy: physics of image formation and microanalysis. Vol. 45. 2013: Springer. 122. Marinello, F., et al., Critical factors in SEM 3D stereo microscopy. Measurement Science and Technology, 2008. 19(6): p. 065705. 123. Myles, T., 3D Imaging with a Scanning Helium Microscope. 2016, University of Newcastle. 124. Roth, A., Vacuum technology. 2012: Elsevier. 125. Moore, J., et al., Introduction to Nanoscience and Nanotechnology. 2008, CRC Press. 126. Niu, S., et al., Excellent Structure-Based Multifunction of Morpho Butterfly Wings: A Review. Journal of Bionic Engineering, 2015. 12(2): p. 170-189.

209 References

127. Williams, C.S. and O.A. Becklund, Introduction to the optical transfer function. 1989: Wiley New York etc. 128. He, B.B., Two-dimensional X-ray diffraction. 2011: John Wiley & Sons. 129. Charland, P., E. El-Khatib, and J. Wolters, The use of deconvolution and total least squares in recovering a radiation detector line spread function. Medical Physics, 1998. 25(2): p. 152-160. 130. Seah, M.P., Resolution parameters for model functions used in surface analysis. Surface and Interface Analysis, 2002. 33(12): p. 950-953. 131. Yoshio, S., et al., Performance Test of Fresnel Zone Plate with 50 nm Outermost Zone Width in Hard X-ray Region. Japanese Journal of Applied Physics, 2005. 44(4R): p. 1994. 132. Burke, K.B., et al., A knife-edge measurement of the beam profile of STXM 5.3.2.2 using a focussed ion beam milled metallic glass. Journal of Electron Spectroscopy and Related Phenomena, 2012. 185(11): p. 453-457. 133. Braun, J., et al., Micrometer-sized nozzles and skimmers for the production of supersonic He atom beams. Review of Scientific Instruments, 1997. 68(8): p. 3001-3009. 134. Hedgeland, H., et al., Anomalous attenuation at low temperatures in high- intensity helium beam sources. Review of Scientific Instruments, 2005. 76(12): p. 123111. 135. Luria, K., W. Christen, and U. Even, Generation and Propagation of Intense Supersonic Beams. The Journal of Physical Chemistry A, 2011. 115(25): p. 7362- 7367. 136. Kingslake, R., R.R. Shannon, and J.C. Wyant, Applied Optics and Optical Engineering. 1983: Academic Press. 137. Roth, G.D., Handbook of Practical Astronomy. 2009: Springer Berlin Heidelberg. 138. de Villiers, G. and E.R. Pike, The Limits of Resolution. 2016: CRC Press. 139. Stewart, J., Optical principles and technology for engineers. 2014: CRC Press. 140. DePonte, D., DP DePonte, SD Kevan, and FS Patton, Rev. Sci. Instrum. 77, 055107 (2006). Rev. Sci. Instrum., 2006. 77: p. 055107. 141. Reisinger, T. and B. Holst, Neutral atom and molecule focusing using a Fresnel zone plate. Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and Phenomena, 2008. 26(6): p. 2374-2379. 142. Carnal, O., et al., Imaging and focusing of atoms by a fresnel zone plate. Physical Review Letters, 1991. 67(23): p. 3231-3234. 143. Carnal, O. and J. Mlynek, Atom Optics with Diffractive Structures. Europhys. News, 1992. 23(8): p. 149-152. 144. Ekstrom, C.R., D.W. Keith, and D.E. Pritchard, Atom optics using microfabricated structures. Applied Physics B, 1992. 54(5): p. 369-374. 145. Mlynek, J., V. Balykin, and P. Meystre, Optics and interferometry with atoms. Applied Physics B, 1992. 54(5): p. 319-320.

210 References

146. Sleator, T., et al., Imaging and focusing of an atomic beam with a large period standing light wave. Applied Physics B, 1992. 54(5): p. 375-379. 147. Adams, C.S., M. Sigel, and J. Mlynek, Atom optics. Physics Reports, 1994. 240(3): p. 143-210. 148. Kaenders, W.G., et al., Imaging with an atomic beam. Nature, 1995. 375: p. 214. 149. Barr, M., et al., A design for a pinhole scanning helium microscope. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, 2014. 340: p. 76-80. 150. Fahy, A., et al., A highly contrasting scanning helium microscope. Review of Scientific Instruments, 2015. 86(2): p. 023704. 151. Barr, M., et al., Unlocking new contrast in a scanning helium microscope. Nature Communications, 2016. 7: p. 10189. 152. Witham, P.J. and E.J. Sanchez, Increased resolution in neutral atom microscopy. Journal of Microscopy, 2012. 248(3): p. 223-227. 153. Eder, S.D., et al., Focusing of a neutral helium beam below one micron. New Journal of Physics, 2012. 14(7): p. 073014. 154. Schofield, N., Progress Towards Helium Beam Microscopy: Focusing Using Optical Potentials, in School of Physics. 2012, Monash University. 155. Salvador Palau, A., G. Bracco, and B. Holst, Theoretical model of the helium zone plate microscope. Physical Review A, 2017. 95(1): p. 013611. 156. Verheijen, M.J., et al., A quantitative description of skimmer interaction in supersonic secondary beams: Calibration of absolute intensities. Chemical Physics, 1984. 85(1): p. 63-71. 157. Campargue, R., Aerodynamic Separation Effect on Gas and Isotope Mixtures Induced by Invasion of the Free Jet Shock Wave Structure. The Journal of Chemical Physics, 1970. 52(4): p. 1795-1802. 158. Even, U., Pulsed Supersonic Beams from High Pressure Source: Simulation Results and Experimental Measurements. Advances in Chemistry, 2014. 2014: p. 11. 159. Greve, M.M., et al., The Beynon Gabor zone plate: a new tool for de Broglie matter waves and hard X-rays? An off axis and focus intensity investigation. Optics Express, 2013. 21(23): p. 28483-28495. 160. Eder, S.D., et al., Zero-order filter for diffractive focusing of de Broglie matter waves. Physical Review A, 2017. 95(2): p. 023618. 161. Eder, S.D., et al., Focusing of a neutral helium beam with a photon-sieve structure. Physical Review A, 2015. 91(4): p. 043608. 162. Skinner, G.K., Imaging with coded-aperture masks. Nuclear Instruments and Methods in Physics Research, 1984. 221(1): p. 33-40. 163. Levin, A., et al., Image and depth from a conventional camera with a coded aperture. ACM Trans. Graph., 2007. 26(3): p. 70. 164. Anderson, D.N., et al., Detection and Location of Gamma-Ray Sources With a Modulating Coded Mask. Technometrics, 2006. 48(2): p. 252-261.

211 References

165. Caroli, E., et al., Coded aperture imaging in X- and gamma-ray astronomy. Space Science Reviews, 1987. 45(3): p. 349-403. 166. Barbara A. J. Lechner, An Improved High Intensity Supersonic Helium Beam Source, in Studying Complex Surface Dynamical Systems Using Helium-3 Spin- Echo Spectroscopy, B.A.J. Lechner, Editor. 2014, Springer International Publishing: Cham. p. 51-69. 167. Eder, S.D., et al., A free jet (supersonic), molecular beam source with automatized, 50 nm precision nozzle-skimmer positioning. Review of Scientific Instruments, 2013. 84(9): p. 093303. 168. Cortenraad, R., et al., Growth, characterisation and surface cleaning procedures for high-purity tungsten single crystals. Journal of Crystal Growth, 2001. 222(1): p. 154-162. 169. Taglauer, E., Surface cleaning using sputtering. Applied Physics A, 1990. 51(3): p. 238-251. 170. Yamada, I., Historical milestones and future prospects of cluster ion beam technology. Applied Surface Science, 2014. 310: p. 77-88.

212 References

213 References

214